Search

Machine learning Filter

What is Linear Regression in Machine Learning

Machine Learning, being a subset of Artificial Intelligence (AI), has been playing a dominant role in our daily lives. Data science engineers and developers working in various domains are widely using machine learning algorithms to make their tasks simpler and life easier. For example, certain machine learning algorithms enable Google Maps to find the fastest route to our destinations, allow Tesla to make driverless cars, help Amazon to generate almost 35% of their annual income, AccuWeather to get the weather forecast of 3.5 million locations weeks in advance, Facebook to automatically detect faces and suggest tags and so on.In statistics and machine learning, linear regression is one of the most popular and well understood algorithms. Most data science enthusiasts and machine learning  fanatics begin their journey with linear regression algorithms. In this article, we will look into how linear regression algorithm works and how it can be efficiently used in your machine learning projects to build better models.Linear Regression is one of the machine learning algorithms where the result is predicted by the use of known parameters which are correlated with the output. It is used to predict values within a continuous range rather than trying to classify them into categories. The known parameters are used to make a continuous and constant slope which is used to predict the unknown or the result.What is a Regression Problem?Majority of the machine learning algorithms fall under the supervised learning category. It is the process where an algorithm is used to predict a result based on the previously entered values and the results generated from them. Suppose we have an input variable ‘x’ and an output variable ‘y’ where y is a function of x (y=f{x}). Supervised learning reads the value of entered variable ‘x’ and the resulting variable ‘y’ so that it can use those results to later predict a highly accurate output data of ‘y’ from the entered value of ‘x’. A regression problem is when the resulting variable contains a real or a continuous value. It tries to draw the line of best fit from the data gathered from a number of points.For example, which of these is a regression problem?How much gas will I spend if I drive for 100 miles?What is the nationality of a person?What is the age of a person?Which is the closest planet to the Sun?Predicting the amount of gas to be spent and the age of a person are regression problems. Predicting nationality is categorical and the closest planet to the Sun is discrete.What is Linear Regression?Let’s say we have a dataset which contains information about the relationship between ‘number of hours studied’ and ‘marks obtained’. A number of students have been observed and their hours of study along with their grades are recorded. This will be our training data. Our goal is to design a model that can predict the marks if number of hours studied is provided. Using the training data, a regression line is obtained which will give minimum error. This linear equation is then used to apply for a new data. That is, if we give the number of hours studied by a student as an input, our model should be able to predict their mark with minimum error.Hypothesis of Linear RegressionThe linear regression model can be represented by the following equation:where,Y is the predicted valueθ₀ is the bias term.θ₁,…,θn are the model parametersx₁, x₂,…,xn are the feature values.The above hypothesis can also be represented byWhere, θ is the model’s parameter vector including the bias term θ₀; x is the feature vector with x₀ =1Y (pred) = b0 + b1*xThe values b0 and b1 must be chosen so that the error is minimum. If sum of squared error is taken as a metric to evaluate the model, then the goal is to obtain a line that best reduces the error.If we don’t square the error, then the positive and negative points will cancel each other out.For a model with one predictor,Exploring ‘b1’If b1 > 0, then x (predictor) and y(target) have a positive relationship. That is an increase in x will increase y.If b1 < 0, then x (predictor) and y(target) have a negative relationship. That is an increase in x will decrease y.Exploring ‘b0’If the model does not include x=0, then the prediction will become meaningless with only b0. For example, we have a dataset that relates height(x) and weight(y). Taking x=0 (that is height as 0), will make the equation have only b0 value which is completely meaningless as in real-time height and weight can never be zero. This resulted due to considering the model values beyond its scope.If the model includes value 0, then ‘b0’ will be the average of all predicted values when x=0. But, setting zero for all the predictor variables is often impossible.The value of b0 guarantees that the residual will have mean zero. If there is no ‘b0’ term, then the regression will be forced to pass over the origin. Both the regression coefficient and prediction will be biased.How does Linear Regression work?Let’s look at a scenario where linear regression might be useful: losing weight. Let us consider that there’s a connection between how many calories you take in and how much you weigh; regression analysis can help you understand that connection. Regression analysis will provide you with a relation which can be visualized into a graph in order to make predictions about your data. For example, if you’ve been putting on weight over the last few years, it can predict how much you’ll weigh in the next ten years if you continue to consume the same amount of calories and burn them at the same rate.The goal of regression analysis is to create a trend line based on the data you have gathered. This then allows you to determine whether other factors apart from the amount of calories consumed affect your weight, such as the number of hours you sleep, work pressure, level of stress, type of exercises you do etc. Before taking into account, we need to look at these factors and attributes and determine whether there is a correlation between them. Linear Regression can then be used to draw a trend line which can then be used to confirm or deny the relationship between attributes. If the test is done over a long time duration, extensive data can be collected and the result can be evaluated more accurately. By the end of this article we will build a model which looks like the below picture i.e, determine a line which best fits the data.How do we determine the best fit line?The best fit line is considered to be the line for which the error between the predicted values and the observed values is minimum. It is also called the regression line and the errors are also known as residuals. The figure shown below shows the residuals. It can be visualized by the vertical lines from the observed data value to the regression line.When to use Linear Regression?Linear Regression’s power lies in its simplicity, which means that it can be used to solve problems across various fields. At first, the data collected from the observations need to be collected and plotted along a line. If the difference between the predicted value and the result is almost the same, we can use linear regression for the problem.Assumptions in linear regressionIf you are planning to use linear regression for your problem then there are some assumptions you need to consider:The relation between the dependent and independent variables should be almost linear.The data is homoscedastic, meaning the variance between the results should not be too much.The results obtained from an observation should not be influenced by the results obtained from the previous observation.The residuals should be normally distributed. This assumption means that the probability density function of the residual values is normally distributed at each independent value.You can determine whether your data meets these conditions by plotting it and then doing a bit of digging into its structure.Few properties of Regression LineHere are a few features a regression line has:Regression passes through the mean of independent variable (x) as well as mean of the dependent variable (y).Regression line minimizes the sum of “Square of Residuals”. That’s why the method of Linear Regression is known as “Ordinary Least Square (OLS)”. We will discuss more in detail about Ordinary Least Square later on.B1 explains the change in Y with a change in x  by one unit. In other words, if we increase the value of ‘x’ it will result in a change in value of Y.Finding a Linear Regression lineLet’s say we want to predict ‘y’ from ‘x’ given in the following table and assume they are correlated as “y=B0+B1∗x”xyPredicted 'y'12Β0+B1∗121Β0+B1∗233Β0+B1∗346Β0+B1∗459Β0+B1∗5611Β0+B1∗6713Β0+B1∗7815Β0+B1∗8917Β0+B1∗91020Β0+B1∗10where,Std. Dev. of x3.02765Std. Dev. of y6.617317Mean of x5.5Mean of y9.7Correlation between x & y0.989938If the Residual Sum of Square (RSS) is differentiated with respect to B0 & B1 and the results equated to zero, we get the following equation:B1 = Correlation * (Std. Dev. of y/ Std. Dev. of x)B0 = Mean(Y) – B1 * Mean(X)Putting values from table 1 into the above equations,B1 = 2.64B0 = -2.2Hence, the least regression equation will become –Y = -2.2 + 2.64*xxY - ActualY - Predicted120.44213.08335.72468.36591161113.6471316.2881518.9291721.56102024.2As there are only 10 data points, the results are not too accurate but if we see the correlation between the predicted and actual line, it has turned out to be very high; both the lines are moving almost together and here is the graph for visualizing our predicted values:Model PerformanceAfter the model is built, if we see that the difference in the values of the predicted and actual data is not much, it is considered to be a good model and can be used to make future predictions. The amount that we consider “not much” entirely depends on the task you want to perform and to what percentage the variation in data can be handled. Here are a few metric tools we can use to calculate error in the model-R – Square (R2)Total Sum of Squares (TSS): total sum of squares (TSS) is a quantity that appears as part of a standard way of presenting results of such an analysis. Sum of squares is a measure of how a data set varies around a central number (like the mean). The Total Sum of Squares tells how much variation there is in the dependent variable.TSS = Σ (Y – Mean[Y])2Residual Sum of Squares (RSS): The residual sum of squares tells you how much of the dependent variable’s variation your model did not explain. It is the sum of the squared differences between the actual Y and the predicted Y.RSS = Σ (Y – f[Y])2(TSS – RSS) measures the amount of variability in the response that is explained by performing the regression.Properties of R2R2 always ranges between 0 to 1.R2 of 0 means that there is no correlation between the dependent and the independent variable.R2 of 1 means the dependent variable can be predicted from the independent variable without any error. An R2 between 0 and 1 indicates the extent to which the dependent variable is predictable. An R2 of 0.20 means that there is 20% of the variance in Y is predictable from X; an R2 of 0.40 means that 40% is predictable; and so on.Root Mean Square Error (RMSE)Root Mean Square Error (RMSE) is the standard deviation of the residuals (prediction errors). The formula for calculating RMSE is:Where N : Total number of observationsWhen standardized observations are used as RMSE inputs, there is a direct relationship with the correlation coefficient. For example, if the correlation coefficient is 1, the RMSE will be 0, because all of the points lie on the regression line (and therefore there are no errors).Mean Absolute Percentage Error (MAPE)There are certain limitations to the use of RMSE, so analysts prefer MAPE over RMSE which gives error in terms of percentages so that different models can be considered for the task and see how they perform. Formula for calculating MAPE can be written as:Where N : Total number of observationsFeature SelectionFeature selection is the automatic selection of attributes for your data that are most relevant to the predictive model you are working on. It seeks to reduce the number of attributes in the dataset by eliminating the features which are not required for the model construction. Feature selection does not totally eliminate an attribute which is considered for the model, rather it mutes that particular characteristic and works with the features which affects the model.Feature selection method aids your mission to create an accurate predictive model. It helps you by choosing features that will give you as good or better accuracy whilst requiring less data. Feature selection methods can be used to identify and remove unnecessary, irrelevant and redundant attributes from the data that do not contribute to the accuracy of the model or may even decrease the accuracy of the model. Having fewer attributes is desirable because it reduces the complexity of the model, and a simpler model is easier to understand, explain and to work with.Feature Selection Algorithms:Filter Method: This method involves assigning scores to individual features and ranking them. The features that have very little to almost no impact are removed from consideration while constructing the model.Wrapper Method: Wrapper method is quite similar to Filter method except the fact that it considers attributes in a group i.e. a number of attributes are taken and checked whether they are having an impact on the model and if not another combination is applied.Embedded Method: Embedded method is the best and most accurate of all the algorithms. It learns the features that affect the model while the model is being constructed and takes into consideration only those features. The most common type of embedded feature selection methods are regularization methods.Cost FunctionCost function helps to figure out the best possible plots which can be used to draw the line of best fit for the data points. As we want to reduce the error of the resulting value we change the process of finding out the actual result to a process which can reduce the error between the predicted value and the actual value.Here, J is the cost function.The above function is made in this format to calculate the error difference between the predicted values and the plotted values. We take the square of the summation of all the data points and divide it by the total number of data points. This cost function J is also called the Mean Squared Error (MSE) function. Using this MSE function we are going to predict values such that the MSE value settles at the minima, reducing the cost function.Gradient DescentGradient Descent is an optimization algorithm that helps machine learning models to find out paths to a minimum value using repeated steps. Gradient descent is used to minimize a function so that it gives the lowest output of that function. This function is called the Loss Function. The loss function shows us how much error is produced by the machine learning model compared to actual results. Our aim should be to lower the cost function as much as possible. One way of achieving a low cost function is by the process of gradient descent. Complexity of some equations makes it difficult to use, partial derivative of the cost function with respect to the considered parameter can provide optimal coefficient value. You may refer to the article on Gradient Descent for Machine Learning.Simple Linear RegressionOptimization is a big part of machine learning and almost every machine learning algorithm has an optimization technique at its core for increased efficiency. Gradient Descent is such an optimization algorithm used to find values of coefficients of a function that minimizes the cost function. Gradient Descent is best applied when the solution cannot be obtained by analytical methods (linear algebra) and must be obtained by an optimization technique.Residual Analysis: Simple linear regression models the relationship between the magnitude of one variable and that of a second—for example, as x increases, y also increases. Or as x increases, y decreases. Correlation is another way to measure how two variables are related. The models done by simple linear regression estimate or try to predict the actual result but most often they deviate from the actual result. Residual analysis is used to calculate by how much the estimated value has deviated from the actual result.Null Hypothesis and p-value: During feature selection, null hypothesis is used to find which attributes will not affect the result of the model. Hypothesis tests are used to test the validity of a claim that is made about a particular attribute of the model. This claim that’s on trial, in essence, is called the null hypothesis. A p-value helps to determine the significance of the results. p-value is a number between 0 and 1 and is interpreted in the following way:A small p-value (less than 0.05) indicates a strong evidence against the null hypothesis, so the null hypothesis is to be rejected.A large p-value (greater than 0.05) indicates weak evidence against the null hypothesis, so the null hypothesis is to be considered.p-value very close to the cut-off (equal to 0.05) is considered to be marginal (could go either way). In this case, the p-value should be provided to the readers so that they can draw their own conclusions.Ordinary Least SquareOrdinary Least Squares (OLS), also known as Ordinary least squares regression or least squared errors regression is a type of linear least squares method for estimating the unknown parameters in a linear regression model. OLS chooses the parameters for a linear function, the goal of which is to minimize the sum of the squares of the difference of the observed variables and the dependent variables i.e. it tries to attain a relationship between them. There are two types of relationships that may occur: linear and curvilinear. A linear relationship is a straight line that is drawn through the central tendency of the points; whereas a curvilinear relationship is a curved line. Association between the variables are depicted by using a scatter plot. The relationship could be positive or negative, and result variation also differs in strength.The advantage of using Ordinary Least Squares regression is that it can be easily interpreted and is highly compatible with recent computers’ built-in algorithms from linear algebra. It can be used to apply to problems with lots of independent variables which can efficiently conveyed to thousands of data points. In Linear Regression, OLS is used to estimate the unknown parameters by creating a model which will minimize the sum of the squared errors between the observed data and the predicted one.Let us simulate some data and look at how the predicted values (Yₑ) differ from the actual value (Y):import pandas as pd import numpy as np from matplotlib import pyplot as plt # Generate 'random' data np.random.seed(0) X = 2.5 * np.random.randn(100) + 1.5   # Array of 100 values with mean = 1.5, stddev = 2.5 res = 0.5 * np.random.randn(100)         # Generate 100 residual terms y = 2 + 0.3 * X + res                   # Actual values of Y # Create pandas dataframe to store our X and y values df = pd.DataFrame(     {'X': X,       'y': y} ) # Show the first five rows of our dataframe df.head()XY05.9101314.71461512.5003932.07623823.9468452.54881137.1022334.61536846.1688953.264107To estimate y using the OLS method, we need to calculate xmean and ymean, the covariance of X and y (xycov), and the variance of X (xvar) before we can determine the values for alpha and beta.# Calculate the mean of X and y xmean = np.mean(X) ymean = np.mean(y) # Calculate the terms needed for the numator and denominator of beta df['xycov'] = (df['X'] - xmean) * (df['y'] - ymean) df['xvar'] = (df['X'] - xmean)**2 # Calculate beta and alpha beta = df['xycov'].sum() / df['xvar'].sum() alpha = ymean - (beta * xmean) print(f'alpha = {alpha}') print(f'beta = {beta}')alpha = 2.0031670124623426 beta = 0.3229396867092763Now that we have an estimate for alpha and beta, we can write our model as Yₑ = 2.003 + 0.323 X, and make predictions:ypred = alpha + beta * XLet’s plot our prediction ypred against the actual values of y, to get a better visual understanding of our model.# Plot regression against actual data plt.figure(figsize=(12, 6)) plt.plot(X, ypred) # regression line plt.plot(X, y, 'ro')   # scatter plot showing actual data plt.title('Actual vs Predicted') plt.xlabel('X') plt.ylabel('y') plt.show()The blue line in the above graph is our line of best fit, Yₑ = 2.003 + 0.323 X.  If you observe the graph carefully, you will notice that there is a linear relationship between X and Y. Using this model, we can predict Y from any values of X. For example, for X = 8,Yₑ = 2.003 + 0.323 (8) = 4.587RegularizationRegularization is a type of regression that is used to decrease the coefficient estimates down to zero. This helps to eliminate the data points that don’t actually represent the true properties of the model, but have appeared by random chance. The process is done by identifying the points which have deviated from the line of best-fit by a large extent. Earlier we saw that to estimate the regression coefficients β in the least squares method, we must minimize the term Residual Sum of Squares (RSS). Let the RSS equation in this case be:The general linear regression model can be expressed using a condensed formula:Here, β=[β0 ,β1, ….. βp]The RSS value will adjust the coefficient, β based on the training data. If the resulting data deviates too much from the training data, then the estimated coefficients won’t generalize well to the future data. This is where regularization comes in and shrinks or regularizes these learned estimates towards zero.Ridge regressionRidge regression is very similar to least squares, except that the Ridge coefficients are estimated by minimizing a different quantity. In particular, the Ridge regression coefficients β are the values that minimize the following quantity:Here, λ is the tuning parameter that decides how much we want to penalize the flexibility of the model. λ controls the relative impact of the two components: RSS and the penalty term. If λ = 0, the Ridge regression will produce a result similar to least squares method. If λ → ∞, all estimated coefficients tend to zero. Ridge regression produces different estimates for different values of λ. The optimal choice of λ is crucial and should be done with cross-validation. The coefficient estimates produced by ridge regression method is also known as the L2 norm.The coefficients generated by Ordinary Least Squares method is independent of scale, which means that if each input variable is multiplied by a constant, the corresponding coefficient will be divided by the same constant, as a result of which the multiplication of the coefficient and the input variables will remain the same. The same is not true for ridge regression and we need to bring the coefficients to the same scale before we perform the process. To standardize the variables, we must subtract their means and divide it by their standard deviations.Lasso RegressionLeast Absolute Shrinkage and Selection Operator (LASSO) regression also shrinks the coefficients by adding a penalty to the sum of squares of the residuals, but the lasso penalty has a slightly different effect. The lasso penalty is the sum of the absolute values of the coefficient vector, which corresponds to its L1 norm. Hence, the lasso estimate is defined by:Similar to ridge regression, the input variables need to be standardized. The lasso penalty makes the solution nonlinear, and there is no closed-form expression for the coefficients as in ridge regression. Instead, the lasso solution is a quadratic programming problem and there are available efficient algorithms that compute the entire path of coefficients that result for different values of λ with the same computational cost as for ridge regression.The lasso penalty had the effect of gradually reducing some coefficients to zero as the regularization increases. For this reason, the lasso can be used for the continuous selection of a subset of features.Linear Regression with multiple variablesLinear regression with multiple variables is also known as "multivariate linear regression". We now introduce notation for equations where we can have any number of input variables.x(i)j=value of feature j in the ith training examplex(i)=the input (features) of the ith training examplem=the number of training examplesn=the number of featuresThe multivariable form of the hypothesis function accommodating these multiple features is as follows:hθ(x)=θ0+θ1x1+θ2x2+θ3x3+⋯+θnxnIn order to develop intuition about this function, we can think about θ0 as the basic price of a house, θ1 as the price per square meter, θ2 as the price per floor, etc. x1 will be the number of square meters in the house, x2 the number of floors, etc.Using the definition of matrix multiplication, our multivariable hypothesis function can be concisely represented as:This is a vectorization of our hypothesis function for one training example; see the lessons on vectorization to learn more.Remark: Note that for convenience reasons in this course we assume x0 (i) =1 for (i∈1,…,m). This allows us to do matrix operations with θ and x. Hence making the two vectors ‘θ’and x(i) match each other element-wise (that is, have the same number of elements: n+1).Multiple Linear RegressionHow is it different?In simple linear regression we use a single independent variable to predict the value of a dependent variable whereas in multiple linear regression two or more independent variables are used to predict the value of a dependent variable. The difference between the two is the number of independent variables. In both cases there is only a single dependent variable.MulticollinearityMulticollinearity tells us the strength of the relationship between independent variables. Multicollinearity is a state of very high intercorrelations or inter-associations among the independent variables. It is therefore a type of disturbance in the data, and if present in the data the statistical inferences made about the data may not be reliable. VIF (Variance Inflation Factor) is used to identify the Multicollinearity. If VIF value is greater than 4, we exclude that variable from our model.There are certain reasons why multicollinearity occurs:It is caused by an inaccurate use of dummy variables.It is caused by the inclusion of a variable which is computed from other variables in the data set.Multicollinearity can also result from the repetition of the same kind of variable.Generally occurs when the variables are highly correlated to each other.Multicollinearity can result in several problems. These problems are as follows:The partial regression coefficient due to multicollinearity may not be estimated precisely. The standard errors are likely to be high.Multicollinearity results in a change in the signs as well as in the magnitudes of the partial regression coefficients from one sample to another sample.Multicollinearity makes it tedious to assess the relative importance of the independent variables in explaining the variation caused by the dependent variable.Iterative ModelsModels should be tested and upgraded again and again for better performance. Multiple iterations allows the model to learn from its previous result and take that into consideration while performing the task again.Making predictions with Linear RegressionLinear Regression can be used to predict the value of an unknown variable using a known variable by the help of a straight line (also called the regression line). The prediction can only be made if it is found that there is a significant correlation between the known and the unknown variable through both a correlation coefficient and a scatterplot.The general procedure for using regression to make good predictions is the following:Research the subject-area so that the model can be built based on the results produced by similar models. This research helps with the subsequent steps.Collect data for appropriate variables which have some correlation with the model.Specify and assess the regression model.Run repeated tests so that the model has more data to work with.To test if the model is good enough observe whether:The scatter plot forms a linear pattern.The correlation coefficient r, has a value above 0.5 or below -0.5. A positive value indicates a positive relationship and a negative value represents a negative relationship.If the correlation coefficient shows a strong relationship between variables but the scatter plot is not linear, the results can be misleading. Examples on how to use linear regression have been shown earlier.Data preparation for Linear RegressionStep 1: Linear AssumptionThe first step for data preparation is checking for the variables which have some sort of linear correlation between the dependent and the independent variables.Step 2: Remove NoiseIt is the process of reducing the number of attributes in the dataset by eliminating the features which have very little to no requirement for the construction of the model.Step 3: Remove CollinearityCollinearity tells us the strength of the relationship between independent variables. If two or more variables are highly collinear, it would not make sense to keep both the variables while evaluating the model and hence we can keep one of them.Step 4: Gaussian DistributionsThe linear regression model will produce more reliable results if the input and output variables have a Gaussian distribution. The Gaussian theorem states that  states that a sample mean from an infinite population is approximately normal, or Gaussian, with mean the same as the underlying population, and variance equal to the population variance divided by the sample size. The approximation improves as the sample size gets large.Step 5: Rescale InputsLinear regression model will produce more reliable predictions if the input variables are rescaled using standardization or normalization.Linear Regression with statsmodelsWe have already discussed OLS method, now we will move on and see how to use the OLS method in the statsmodels library. For this we will be using the popular advertising dataset. Here, we will only be looking at the TV variable and explore whether spending on TV advertising can predict the number of sales for the product. Let’s start by importing this csv file as a pandas dataframe using read_csv():# Import and display first five rows of advertising dataset advert = pd.read_csv('advertising.csv') advert.head()TVRadioNewspaperSales0230.137.869.222.1144.539.345.110.4217.245.969.312.03151.541.358.516.54180.810.858.417.9Now we will use statsmodels’ OLS function to initialize simple linear regression model. It will take the formula y ~ X, where X is the predictor variable (TV advertising costs) and y is the output variable (Sales). Then, we will fit the model by calling the OLS object’s fit() method.import statsmodels.formula.api as smf # Initialise and fit linear regression model using `statsmodels` model = smf.ols('Sales ~ TV', data=advert) model = model.fit()Once we have fit the simple regression model, we can predict the values of sales based on the equation we just derived using the .predict method and also visualise our regression model by plotting sales_pred against the TV advertising costs to find the line of best fit.# Predict values sales_pred = model.predict() # Plot regression against actual data plt.figure(figsize=(12, 6)) plt.plot(advert['TV'], advert['Sales'], 'o')       # scatter plot showing actual data plt.plot(advert['TV'], sales_pred, 'r', linewidth=2)   # regression line plt.xlabel('TV Advertising Costs') plt.ylabel('Sales') plt.title('TV vs Sales') plt.show()In the above graph, if you notice you will see that there is a positive linear relationship between TV advertising costs and Sales. You may also summarize by saying that spending more on TV advertising predicts a higher number of sales.Linear Regression with scikit-learnLet us learn to implement linear regression models using sklearn. For this model as well, we will continue to use the advertising dataset but now we will use two predictor variables to create a multiple linear regression model. Yₑ = α + β₁X₁ + β₂X₂ + … + βₚXₚ, where p is the number of predictors.In our example, we will be predicting Sales using the variables TV and Radio i.e. our model can be written as:Sales = α + β₁*TV + β₂*Radiofrom sklearn.linear_model import LinearRegression # Build linear regression model using TV and Radio as predictors # Split data into predictors X and output Y predictors = ['TV', 'Radio'] X = advert[predictors] y = advert['Sales'] # Initialise and fit model lm = LinearRegression() model = lm.fit(X, y) print(f'alpha = {model.intercept_}') print(f'betas = {model.coef_}')alpha = 4.630879464097768 betas = [0.05444896 0.10717457]model.predict(X)Now that we have fit a multiple linear regression model to our data, we can predict sales from any combination of TV and Radio advertising costs. For example, you want to know how many sales we would make if we invested $600 in TV advertising and $300 in Radio advertising. You can simply find it out by:new_X = [[600, 300]] print(model.predict(new_X))[69.4526273]We get the output as 69.45 which means if we invest $600 on TV and $300 on Radio advertising, we can expect to sell 69 units approximately.SummaryLet us sum up what we have covered in this article so far —How to understand a regression problemWhat is linear regression and how it worksOrdinary Least Square method and RegularizationImplementing Linear Regression in Python using statsmodel and sklearn libraryWe have discussed about a couple of ways to implement linear regression and build efficient models for certain business problems. If you are inspired by the opportunities provided by machine learning, enrol in our  Data Science and Machine Learning Courses for more lucrative career options in this landscape.

What is Linear Regression in Machine Learning

8961
What is Linear Regression in Machine Learning

Machine Learning, being a subset of Artificial Intelligence (AI), has been playing a dominant role in our daily lives. Data science engineers and developers working in various domains are widely using machine learning algorithms to make their tasks simpler and life easier. For example, certain machine learning algorithms enable Google Maps to find the fastest route to our destinations, allow Tesla to make driverless cars, help Amazon to generate almost 35% of their annual income, AccuWeather to get the weather forecast of 3.5 million locations weeks in advance, Facebook to automatically detect faces and suggest tags and so on.

In statistics and machine learning, linear regression is one of the most popular and well understood algorithms. Most data science enthusiasts and machine learning  fanatics begin their journey with linear regression algorithms. In this article, we will look into how linear regression algorithm works and how it can be efficiently used in your machine learning projects to build better models.

Linear Regression is one of the machine learning algorithms where the result is predicted by the use of known parameters which are correlated with the output. It is used to predict values within a continuous range rather than trying to classify them into categories. The known parameters are used to make a continuous and constant slope which is used to predict the unknown or the result.

What is a Regression Problem?

Majority of the machine learning algorithms fall under the supervised learning category. It is the process where an algorithm is used to predict a result based on the previously entered values and the results generated from them. Suppose we have an input variable ‘x’ and an output variable ‘y’ where y is a function of x (y=f{x}). Supervised learning reads the value of entered variable ‘x’ and the resulting variable ‘y’ so that it can use those results to later predict a highly accurate output data of ‘y’ from the entered value of ‘x’. A regression problem is when the resulting variable contains a real or a continuous value. It tries to draw the line of best fit from the data gathered from a number of points.

What is a Regression Problem?

For example, which of these is a regression problem?

  • How much gas will I spend if I drive for 100 miles?
  • What is the nationality of a person?
  • What is the age of a person?
  • Which is the closest planet to the Sun?

Predicting the amount of gas to be spent and the age of a person are regression problems. Predicting nationality is categorical and the closest planet to the Sun is discrete.

What is Linear Regression?

Let’s say we have a dataset which contains information about the relationship between ‘number of hours studied’ and ‘marks obtained’. A number of students have been observed and their hours of study along with their grades are recorded. This will be our training data. Our goal is to design a model that can predict the marks if number of hours studied is provided. Using the training data, a regression line is obtained which will give minimum error. This linear equation is then used to apply for a new data. That is, if we give the number of hours studied by a student as an input, our model should be able to predict their mark with minimum error.

Hypothesis of Linear Regression

The linear regression model can be represented by the following equation:

The linear regression model equation

where,

Y is the predicted value

θ₀ is the bias term.

θ₁,…,θn are the model parameters

x₁, x₂,…,xn are the feature values.

The above hypothesis can also be represented by

The above hypothesis

Where, θ is the model’s parameter vector including the bias term θ₀; x is the feature vector with x₀ =1

Y (pred) = b0 + b1*x

The values b0 and b1 must be chosen so that the error is minimum. If sum of squared error is taken as a metric to evaluate the model, then the goal is to obtain a line that best reduces the error.

Error Calculation in Linear Regression

If we don’t square the error, then the positive and negative points will cancel each other out.

For a model with one predictor,

Intercept Calculation in Linear Regression

Coefficient Formula in Linear Regression

Exploring ‘b1

If b1 > 0, then x (predictor) and y(target) have a positive relationship. That is an increase in x will increase y.

If b1 < 0, then x (predictor) and y(target) have a negative relationship. That is an increase in x will decrease y.

Exploring ‘b0

If the model does not include x=0, then the prediction will become meaningless with only b0. For example, we have a dataset that relates height(x) and weight(y). Taking x=0 (that is height as 0), will make the equation have only b0 value which is completely meaningless as in real-time height and weight can never be zero. This resulted due to considering the model values beyond its scope.

If the model includes value 0, then ‘b0’ will be the average of all predicted values when x=0. But, setting zero for all the predictor variables is often impossible.

The value of b0 guarantees that the residual will have mean zero. If there is no ‘b0’ term, then the regression will be forced to pass over the origin. Both the regression coefficient and prediction will be biased.

How does Linear Regression work?

Let’s look at a scenario where linear regression might be useful: losing weight. Let us consider that there’s a connection between how many calories you take in and how much you weigh; regression analysis can help you understand that connection. Regression analysis will provide you with a relation which can be visualized into a graph in order to make predictions about your data. For example, if you’ve been putting on weight over the last few years, it can predict how much you’ll weigh in the next ten years if you continue to consume the same amount of calories and burn them at the same rate.

The goal of regression analysis is to create a trend line based on the data you have gathered. This then allows you to determine whether other factors apart from the amount of calories consumed affect your weight, such as the number of hours you sleep, work pressure, level of stress, type of exercises you do etc. Before taking into account, we need to look at these factors and attributes and determine whether there is a correlation between them. Linear Regression can then be used to draw a trend line which can then be used to confirm or deny the relationship between attributes. If the test is done over a long time duration, extensive data can be collected and the result can be evaluated more accurately. By the end of this article we will build a model which looks like the below picture i.e, determine a line which best fits the data.

How does Linear Regression work?

How do we determine the best fit line?

The best fit line is considered to be the line for which the error between the predicted values and the observed values is minimum. It is also called the regression line and the errors are also known as residuals. The figure shown below shows the residuals. It can be visualized by the vertical lines from the observed data value to the regression line.

How do we determine the best fit line?

When to use Linear Regression?

Linear Regression’s power lies in its simplicity, which means that it can be used to solve problems across various fields. At first, the data collected from the observations need to be collected and plotted along a line. If the difference between the predicted value and the result is almost the same, we can use linear regression for the problem.

Assumptions in linear regression

If you are planning to use linear regression for your problem then there are some assumptions you need to consider:

  • The relation between the dependent and independent variables should be almost linear.
  • The data is homoscedastic, meaning the variance between the results should not be too much.
  • The results obtained from an observation should not be influenced by the results obtained from the previous observation.
  • The residuals should be normally distributed. This assumption means that the probability density function of the residual values is normally distributed at each independent value.

You can determine whether your data meets these conditions by plotting it and then doing a bit of digging into its structure.

Few properties of Regression Line

Here are a few features a regression line has:

  • Regression passes through the mean of independent variable (x) as well as mean of the dependent variable (y).
  • Regression line minimizes the sum of “Square of Residuals”. That’s why the method of Linear Regression is known as “Ordinary Least Square (OLS)”. We will discuss more in detail about Ordinary Least Square later on.
  • B1 explains the change in Y with a change in x  by one unit. In other words, if we increase the value of ‘x’ it will result in a change in value of Y.

Finding a Linear Regression line

Let’s say we want to predict ‘y’ from ‘x’ given in the following table and assume they are correlated as “y=B0+B1∗x”

xyPredicted 'y'
12Β0+B1∗1
21Β0+B1∗2
33Β0+B1∗3
46Β0+B1∗4
59Β0+B1∗5
611Β0+B1∗6
713Β0+B1∗7
815Β0+B1∗8
917Β0+B1∗9
1020Β0+B1∗10

where,

Std. Dev. of x3.02765
Std. Dev. of y6.617317
Mean of x5.5
Mean of y9.7
Correlation between x & y0.989938

If the Residual Sum of Square (RSS) is differentiated with respect to B0 & B1 and the results equated to zero, we get the following equation:

B1 = Correlation * (Std. Dev. of y/ Std. Dev. of x)

B0 = Mean(Y) – B1 * Mean(X)

Putting values from table 1 into the above equations,

B1 = 2.64

B0 = -2.2

Hence, the least regression equation will become –

Y = -2.2 + 2.64*x

xY - ActualY - Predicted
120.44
213.08
335.72
468.36
5911
61113.64
71316.28
81518.92
91721.56
102024.2

As there are only 10 data points, the results are not too accurate but if we see the correlation between the predicted and actual line, it has turned out to be very high; both the lines are moving almost together and here is the graph for visualizing our predicted values:

Finding a Linear Regression line

Model Performance

After the model is built, if we see that the difference in the values of the predicted and actual data is not much, it is considered to be a good model and can be used to make future predictions. The amount that we consider “not much” entirely depends on the task you want to perform and to what percentage the variation in data can be handled. Here are a few metric tools we can use to calculate error in the model-

R – Square (R2)

Model Performance

Total Sum of Squares (TSS): total sum of squares (TSS) is a quantity that appears as part of a standard way of presenting results of such an analysis. Sum of squares is a measure of how a data set varies around a central number (like the mean). The Total Sum of Squares tells how much variation there is in the dependent variable.

TSS = Σ (Y – Mean[Y])2

Residual Sum of Squares (RSS): The residual sum of squares tells you how much of the dependent variable’s variation your model did not explain. It is the sum of the squared differences between the actual Y and the predicted Y.

RSS = Σ (Y – f[Y])2

(TSS – RSS) measures the amount of variability in the response that is explained by performing the regression.

Properties of R2

  • R2 always ranges between 0 to 1.
  • R2 of 0 means that there is no correlation between the dependent and the independent variable.
  • R2 of 1 means the dependent variable can be predicted from the independent variable without any error. 
  • An R2 between 0 and 1 indicates the extent to which the dependent variable is predictable. An R2 of 0.20 means that there is 20% of the variance in Y is predictable from X; an R2 of 0.40 means that 40% is predictable; and so on.

Root Mean Square Error (RMSE)

Root Mean Square Error (RMSE) is the standard deviation of the residuals (prediction errors). The formula for calculating RMSE is:

Root Mean Square Error (RMSE)

Where N : Total number of observations

When standardized observations are used as RMSE inputs, there is a direct relationship with the correlation coefficient. For example, if the correlation coefficient is 1, the RMSE will be 0, because all of the points lie on the regression line (and therefore there are no errors).

Mean Absolute Percentage Error (MAPE)

There are certain limitations to the use of RMSE, so analysts prefer MAPE over RMSE which gives error in terms of percentages so that different models can be considered for the task and see how they perform. Formula for calculating MAPE can be written as:

Mean Absolute Percentage Error (MAPE)

Where N : Total number of observations

Feature Selection

Feature selection is the automatic selection of attributes for your data that are most relevant to the predictive model you are working on. It seeks to reduce the number of attributes in the dataset by eliminating the features which are not required for the model construction. Feature selection does not totally eliminate an attribute which is considered for the model, rather it mutes that particular characteristic and works with the features which affects the model.

Feature selection method aids your mission to create an accurate predictive model. It helps you by choosing features that will give you as good or better accuracy whilst requiring less data. Feature selection methods can be used to identify and remove unnecessary, irrelevant and redundant attributes from the data that do not contribute to the accuracy of the model or may even decrease the accuracy of the model. Having fewer attributes is desirable because it reduces the complexity of the model, and a simpler model is easier to understand, explain and to work with.

Feature Selection Algorithms:

  • Filter Method: This method involves assigning scores to individual features and ranking them. The features that have very little to almost no impact are removed from consideration while constructing the model.
  • Wrapper Method: Wrapper method is quite similar to Filter method except the fact that it considers attributes in a group i.e. a number of attributes are taken and checked whether they are having an impact on the model and if not another combination is applied.
  • Embedded Method: Embedded method is the best and most accurate of all the algorithms. It learns the features that affect the model while the model is being constructed and takes into consideration only those features. The most common type of embedded feature selection methods are regularization methods.

Cost Function

Cost function helps to figure out the best possible plots which can be used to draw the line of best fit for the data points. As we want to reduce the error of the resulting value we change the process of finding out the actual result to a process which can reduce the error between the predicted value and the actual value.

Cost Function in Linear Regression

Here, J is the cost function.

The above function is made in this format to calculate the error difference between the predicted values and the plotted values. We take the square of the summation of all the data points and divide it by the total number of data points. This cost function J is also called the Mean Squared Error (MSE) function. Using this MSE function we are going to predict values such that the MSE value settles at the minima, reducing the cost function.

Gradient Descent

Gradient Descent is an optimization algorithm that helps machine learning models to find out paths to a minimum value using repeated steps. Gradient descent is used to minimize a function so that it gives the lowest output of that function. This function is called the Loss Function. The loss function shows us how much error is produced by the machine learning model compared to actual results. Our aim should be to lower the cost function as much as possible. One way of achieving a low cost function is by the process of gradient descent. Complexity of some equations makes it difficult to use, partial derivative of the cost function with respect to the considered parameter can provide optimal coefficient value. You may refer to the article on Gradient Descent for Machine Learning.

Simple Linear Regression

Optimization is a big part of machine learning and almost every machine learning algorithm has an optimization technique at its core for increased efficiency. Gradient Descent is such an optimization algorithm used to find values of coefficients of a function that minimizes the cost function. Gradient Descent is best applied when the solution cannot be obtained by analytical methods (linear algebra) and must be obtained by an optimization technique.

Residual Analysis: Simple linear regression models the relationship between the magnitude of one variable and that of a second—for example, as x increases, y also increases. Or as x increases, y decreases. Correlation is another way to measure how two variables are related. The models done by simple linear regression estimate or try to predict the actual result but most often they deviate from the actual result. Residual analysis is used to calculate by how much the estimated value has deviated from the actual result.

Null Hypothesis and p-value: During feature selection, null hypothesis is used to find which attributes will not affect the result of the model. Hypothesis tests are used to test the validity of a claim that is made about a particular attribute of the model. This claim that’s on trial, in essence, is called the null hypothesis. A p-value helps to determine the significance of the results. p-value is a number between 0 and 1 and is interpreted in the following way:

  • A small p-value (less than 0.05) indicates a strong evidence against the null hypothesis, so the null hypothesis is to be rejected.
  • A large p-value (greater than 0.05) indicates weak evidence against the null hypothesis, so the null hypothesis is to be considered.
  • p-value very close to the cut-off (equal to 0.05) is considered to be marginal (could go either way). In this case, the p-value should be provided to the readers so that they can draw their own conclusions.

Ordinary Least Square

Ordinary Least Squares (OLS), also known as Ordinary least squares regression or least squared errors regression is a type of linear least squares method for estimating the unknown parameters in a linear regression model. OLS chooses the parameters for a linear function, the goal of which is to minimize the sum of the squares of the difference of the observed variables and the dependent variables i.e. it tries to attain a relationship between them. 

There are two types of relationships that may occur: linear and curvilinear. A linear relationship is a straight line that is drawn through the central tendency of the points; whereas a curvilinear relationship is a curved line. Association between the variables are depicted by using a scatter plot. The relationship could be positive or negative, and result variation also differs in strength.

The advantage of using Ordinary Least Squares regression is that it can be easily interpreted and is highly compatible with recent computers’ built-in algorithms from linear algebra. It can be used to apply to problems with lots of independent variables which can efficiently conveyed to thousands of data points. In Linear Regression, OLS is used to estimate the unknown parameters by creating a model which will minimize the sum of the squared errors between the observed data and the predicted one.

Let us simulate some data and look at how the predicted values (Yₑ) differ from the actual value (Y):

import pandas as pd
import numpy as np
from matplotlib import pyplot as plt

# Generate 'random' data
np.random.seed(0)
X = 2.5 * np.random.randn(100) + 1.5         # Array of 100 values with mean = 1.5, stddev = 2.5
res = 0.5 * np.random.randn(100)         # Generate 100 residual terms
y = 2 + 0.3 * X + res                    # Actual values of Y

# Create pandas dataframe to store our X and y values
df = pd.DataFrame(
    {'X': X,
      'y': y}
)

# Show the first five rows of our dataframe
df.head()

XY
05.9101314.714615
12.5003932.076238
23.9468452.548811
37.1022334.615368
46.1688953.264107

To estimate y using the OLS method, we need to calculate xmean and ymean, the covariance of X and y (xycov), and the variance of X (xvar) before we can determine the values for alpha and beta.

# Calculate the mean of X and y
xmean = np.mean(X)
ymean = np.mean(y)

# Calculate the terms needed for the numator and denominator of beta
df['xycov'] = (df['X'] - xmean) * (df['y'] - ymean)
df['xvar'] = (df['X'] - xmean)**2

# Calculate beta and alpha
beta = df['xycov'].sum() / df['xvar'].sum()
alpha = ymean - (beta * xmean)
print(f'alpha = {alpha}')
print(f'beta = {beta}')
alpha = 2.0031670124623426
beta = 0.3229396867092763

Now that we have an estimate for alpha and beta, we can write our model as Yₑ = 2.003 + 0.323 X, and make predictions:

ypred = alpha + beta * X

Let’s plot our prediction ypred against the actual values of y, to get a better visual understanding of our model.

# Plot regression against actual data
plt.figure(figsize=(12, 6))
plt.plot(X, ypred) # regression line
plt.plot(X, y, 'ro')   # scatter plot showing actual data
plt.title('Actual vs Predicted')
plt.xlabel('X')
plt.ylabel('y')

plt.show()

The blue line in the above graph is our line of best fit

The blue line in the above graph is our line of best fit, Yₑ = 2.003 + 0.323 X.  If you observe the graph carefully, you will notice that there is a linear relationship between X and Y. Using this model, we can predict Y from any values of X. For example, for X = 8,

Yₑ = 2.003 + 0.323 (8) = 4.587

Regularization

Regularization is a type of regression that is used to decrease the coefficient estimates down to zero. This helps to eliminate the data points that don’t actually represent the true properties of the model, but have appeared by random chance. The process is done by identifying the points which have deviated from the line of best-fit by a large extent. Earlier we saw that to estimate the regression coefficients β in the least squares method, we must minimize the term Residual Sum of Squares (RSS). Let the RSS equation in this case be:

Regularization in Linear Regression

The general linear regression model can be expressed using a condensed formula:

expressed using a condensed formula

Here, β=[β01, ….. βp]

The RSS value will adjust the coefficient, β based on the training data. If the resulting data deviates too much from the training data, then the estimated coefficients won’t generalize well to the future data. This is where regularization comes in and shrinks or regularizes these learned estimates towards zero.

Ridge regression

Ridge regression is very similar to least squares, except that the Ridge coefficients are estimated by minimizing a different quantity. In particular, the Ridge regression coefficients β are the values that minimize the following quantity:

Ridge regression in Linear Regression

Here, λ is the tuning parameter that decides how much we want to penalize the flexibility of the model. λ controls the relative impact of the two components: RSS and the penalty term. If λ = 0, the Ridge regression will produce a result similar to least squares method. If λ → ∞, all estimated coefficients tend to zero. Ridge regression produces different estimates for different values of λ. The optimal choice of λ is crucial and should be done with cross-validation. The coefficient estimates produced by ridge regression method is also known as the L2 norm.

The coefficients generated by Ordinary Least Squares method is independent of scale, which means that if each input variable is multiplied by a constant, the corresponding coefficient will be divided by the same constant, as a result of which the multiplication of the coefficient and the input variables will remain the same. The same is not true for ridge regression and we need to bring the coefficients to the same scale before we perform the process. To standardize the variables, we must subtract their means and divide it by their standard deviations.

Lasso Regression

Least Absolute Shrinkage and Selection Operator (LASSO) regression also shrinks the coefficients by adding a penalty to the sum of squares of the residuals, but the lasso penalty has a slightly different effect. The lasso penalty is the sum of the absolute values of the coefficient vector, which corresponds to its L1 norm. Hence, the lasso estimate is defined by:

Lasso Regression in Linear Regression

Similar to ridge regression, the input variables need to be standardized. The lasso penalty makes the solution nonlinear, and there is no closed-form expression for the coefficients as in ridge regression. Instead, the lasso solution is a quadratic programming problem and there are available efficient algorithms that compute the entire path of coefficients that result for different values of λ with the same computational cost as for ridge regression.

The lasso penalty had the effect of gradually reducing some coefficients to zero as the regularization increases. For this reason, the lasso can be used for the continuous selection of a subset of features.

Linear Regression with multiple variables

Linear regression with multiple variables is also known as "multivariate linear regression". We now introduce notation for equations where we can have any number of input variables.

x(i)j=value of feature j in the ith training example

x(i)=the input (features) of the ith training example

m=the number of training examples

n=the number of features

The multivariable form of the hypothesis function accommodating these multiple features is as follows:

hθ(x)=θ01x12x23x3+⋯+θnxn

In order to develop intuition about this function, we can think about θ0 as the basic price of a house, θ1 as the price per square meter, θ2 as the price per floor, etc. x1 will be the number of square meters in the house, x2 the number of floors, etc.

Using the definition of matrix multiplication, our multivariable hypothesis function can be concisely represented as:

Linear Regression with multiple variables

This is a vectorization of our hypothesis function for one training example; see the lessons on vectorization to learn more.

Remark: Note that for convenience reasons in this course we assume x0 (i) =1 for (i∈1,…,m). This allows us to do matrix operations with θ and x. Hence making the two vectors ‘θ’and x(i) match each other element-wise (that is, have the same number of elements: n+1).

Multiple Linear Regression

How is it different?

In simple linear regression we use a single independent variable to predict the value of a dependent variable whereas in multiple linear regression two or more independent variables are used to predict the value of a dependent variable. The difference between the two is the number of independent variables. In both cases there is only a single dependent variable.

Multicollinearity

Multicollinearity tells us the strength of the relationship between independent variables. Multicollinearity is a state of very high intercorrelations or inter-associations among the independent variables. It is therefore a type of disturbance in the data, and if present in the data the statistical inferences made about the data may not be reliable. VIF (Variance Inflation Factor) is used to identify the Multicollinearity. If VIF value is greater than 4, we exclude that variable from our model.

There are certain reasons why multicollinearity occurs:

  • It is caused by an inaccurate use of dummy variables.
  • It is caused by the inclusion of a variable which is computed from other variables in the data set.
  • Multicollinearity can also result from the repetition of the same kind of variable.
  • Generally occurs when the variables are highly correlated to each other.

Multicollinearity can result in several problems. These problems are as follows:

  • The partial regression coefficient due to multicollinearity may not be estimated precisely. The standard errors are likely to be high.
  • Multicollinearity results in a change in the signs as well as in the magnitudes of the partial regression coefficients from one sample to another sample.
  • Multicollinearity makes it tedious to assess the relative importance of the independent variables in explaining the variation caused by the dependent variable.

Iterative Models

Models should be tested and upgraded again and again for better performance. Multiple iterations allows the model to learn from its previous result and take that into consideration while performing the task again.

Making predictions with Linear Regression

Linear Regression can be used to predict the value of an unknown variable using a known variable by the help of a straight line (also called the regression line). The prediction can only be made if it is found that there is a significant correlation between the known and the unknown variable through both a correlation coefficient and a scatterplot.

The general procedure for using regression to make good predictions is the following:

  • Research the subject-area so that the model can be built based on the results produced by similar models. This research helps with the subsequent steps.
  • Collect data for appropriate variables which have some correlation with the model.
  • Specify and assess the regression model.
  • Run repeated tests so that the model has more data to work with.

To test if the model is good enough observe whether:

  • The scatter plot forms a linear pattern.
  • The correlation coefficient r, has a value above 0.5 or below -0.5. A positive value indicates a positive relationship and a negative value represents a negative relationship.

If the correlation coefficient shows a strong relationship between variables but the scatter plot is not linear, the results can be misleading. Examples on how to use linear regression have been shown earlier.

Data preparation for Linear Regression

Step 1: Linear Assumption

The first step for data preparation is checking for the variables which have some sort of linear correlation between the dependent and the independent variables.

Step 2: Remove Noise

It is the process of reducing the number of attributes in the dataset by eliminating the features which have very little to no requirement for the construction of the model.

Step 3: Remove Collinearity

Collinearity tells us the strength of the relationship between independent variables. If two or more variables are highly collinear, it would not make sense to keep both the variables while evaluating the model and hence we can keep one of them.

Step 4: Gaussian Distributions

The linear regression model will produce more reliable results if the input and output variables have a Gaussian distribution. The Gaussian theorem states that  states that a sample mean from an infinite population is approximately normal, or Gaussian, with mean the same as the underlying population, and variance equal to the population variance divided by the sample size. The approximation improves as the sample size gets large.

Step 5: Rescale Inputs

Linear regression model will produce more reliable predictions if the input variables are rescaled using standardization or normalization.

Linear Regression with statsmodels

We have already discussed OLS method, now we will move on and see how to use the OLS method in the statsmodels library. For this we will be using the popular advertising dataset. Here, we will only be looking at the TV variable and explore whether spending on TV advertising can predict the number of sales for the product. Let’s start by importing this csv file as a pandas dataframe using read_csv():

# Import and display first five rows of advertising dataset
advert = pd.read_csv('advertising.csv')
advert.head()

TVRadioNewspaperSales
0230.137.869.222.1
144.539.345.110.4
217.245.969.312.0
3151.541.358.516.5
4180.810.858.417.9

Now we will use statsmodels’ OLS function to initialize simple linear regression model. It will take the formula y ~ X, where X is the predictor variable (TV advertising costs) and y is the output variable (Sales). Then, we will fit the model by calling the OLS object’s fit() method.

import statsmodels.formula.api as smf

# Initialise and fit linear regression model using `statsmodels`
model = smf.ols('Sales ~ TV', data=advert)
model = model.fit()

Once we have fit the simple regression model, we can predict the values of sales based on the equation we just derived using the .predict method and also visualise our regression model by plotting sales_pred against the TV advertising costs to find the line of best fit.

# Predict values
sales_pred = model.predict()

# Plot regression against actual data
plt.figure(figsize=(12, 6))
plt.plot(advert['TV'], advert['Sales'], 'o')       # scatter plot showing actual data
plt.plot(advert['TV'], sales_pred, 'r', linewidth=2)   # regression line
plt.xlabel('TV Advertising Costs')
plt.ylabel('Sales')
plt.title('TV vs Sales')

plt.show()

Linear Regression with statsmodels

In the above graph, if you notice you will see that there is a positive linear relationship between TV advertising costs and Sales. You may also summarize by saying that spending more on TV advertising predicts a higher number of sales.

Linear Regression with scikit-learn

Let us learn to implement linear regression models using sklearn. For this model as well, we will continue to use the advertising dataset but now we will use two predictor variables to create a multiple linear regression model. 

Yₑ = α + β₁X₁ + β₂X₂ + … + βₚXₚ, where p is the number of predictors.

In our example, we will be predicting Sales using the variables TV and Radio i.e. our model can be written as:

Sales = α + β₁*TV + β₂*Radio

from sklearn.linear_model import LinearRegression

# Build linear regression model using TV and Radio as predictors
# Split data into predictors X and output Y
predictors = ['TV', 'Radio']
X = advert[predictors]
y = advert['Sales']

# Initialise and fit model
lm = LinearRegression()
model = lm.fit(X, y)
print(f'alpha = {model.intercept_}')
print(f'betas = {model.coef_}')
alpha = 4.630879464097768
betas = [0.05444896 0.10717457]
model.predict(X)

Linear Regression with scikit-learn

Now that we have fit a multiple linear regression model to our data, we can predict sales from any combination of TV and Radio advertising costs. For example, you want to know how many sales we would make if we invested $600 in TV advertising and $300 in Radio advertising. You can simply find it out by:

new_X = [[600, 300]]
print(model.predict(new_X))
[69.4526273]

We get the output as 69.45 which means if we invest $600 on TV and $300 on Radio advertising, we can expect to sell 69 units approximately.

Summary

Let us sum up what we have covered in this article so far —

  • How to understand a regression problem
  • What is linear regression and how it works
  • Ordinary Least Square method and Regularization
  • Implementing Linear Regression in Python using statsmodel and sklearn library

We have discussed about a couple of ways to implement linear regression and build efficient models for certain business problems. If you are inspired by the opportunities provided by machine learning, enrol in our  Data Science and Machine Learning Courses for more lucrative career options in this landscape.

Priyankur

Priyankur Sarkar

Data Science Enthusiast

Priyankur Sarkar loves to play with data and get insightful results out of it, then turn those data insights and results in business growth. He is an electronics engineer with a versatile experience as an individual contributor and leading teams, and has actively worked towards building Machine Learning capabilities for organizations.

Join the Discussion

Your email address will not be published. Required fields are marked *

Suggested Blogs

Types of Probability Distributions Every Data Science Expert Should know

Data Science has become one of the most popular interdisciplinary fields. It uses scientific approaches, methods, algorithms, and operations to obtain facts and insights from unstructured, semi-structured, and structured datasets. Organizations use these collected facts and insights for efficient production, business growth, and to predict user requirements. Probability distribution plays a significant role in performing data analysis equipping a dataset for training a model. In this article, you will learn about the types of Probability Distribution, random variables, types of discrete distributions, and continuous distribution.  What is Probability Distribution? A Probability Distribution is a statistical method that determines all the probable values and possibilities that a random variable can deliver from a particular range. This range of values will have a lower bound and an upper bound, which we call the minimum and the maximum possible values.  Various factors on which plotting of a value depends are standard deviation, mean (or average), skewness, and kurtosis. All of these play a significant role in Data science as well. We can use probability distribution in physics, engineering, finance, data analysis, machine learning, etc. Significance of Probability distributions in Data Science In a way, most of the data science and machine learning operations are dependent on several assumptions about the probability of your data. Probability distribution allows a skilled data analyst to recognize and comprehend patterns from large data sets; that is, otherwise, entirely random variables and values. Thus, it makes probability distribution a toolkit based on which we can summarize a large data set. The density function and distribution techniques can also help in plotting data, thus supporting data analysts to visualize data and extract meaning. General Properties of Probability Distributions Probability distribution determines the likelihood of any outcome. The mathematical expression takes a specific value of x and shows the possibility of a random variable with p(x). Some general properties of the probability distribution are – The total of all probabilities for any possible value becomes equal to 1. In a probability distribution, the possibility of finding any specific value or a range of values must lie between 0 and 1. Probability distributions tell us the dispersal of the values from the random variable. Consequently, the type of variable also helps determine the type of probability distribution.Common Data Types Before jumping directly into explaining the different probability distributions, let us first understand the different types of probability distributions or the main categories of the probability distribution. Data analysts and data engineers have to deal with a broad spectrum of data, such as text, numerical, image, audio, voice, and many more. Each of these have a specific means to be represented and analyzed. Data in a probability distribution can either be discrete or continuous. Numerical data especially takes one of the two forms. Discrete data: They take specific values where the outcome of the data remains fixed. Like, for example, the consequence of rolling two dice or the number of overs in a T-20 match. In the first case, the result lies between 2 and 12. In the second case, the event will be less than 20. Different types of discrete distributions that use discrete data are: Binomial Distribution Hypergeometric Distribution Geometric Distribution Poisson Distribution Negative Binomial Distribution Multinomial Distribution  Continuous data: It can obtain any value irrespective of bound or limit. Example: weight, height, any trigonometric value, age, etc. Different types of continuous distributions that use continuous data are: Beta distribution Cauchy distribution Exponential distribution Gamma distribution Logistic distribution Weibull distribution Types of Probability Distribution explained Here are some of the popular types of Probability distributions used by data science professionals. (Try all the code using Jupyter Notebook) Normal Distribution: It is also known as Gaussian distribution. It is one of the simplest types of continuous distribution. This probability distribution is symmetrical around its mean value. It also shows that data at close proximity of the mean is frequently occurring, compared to data that is away from it. Here, mean = 0, variance = finite valueHere, you can see 0 at the center is the Normal Distribution for different mean and variance values. Here is a code example showing the use of Normal Distribution: from scipy.stats import norm  import matplotlib.pyplot as mpl  import numpy as np  def normalDist() -> None:      fig, ax = mpl.subplots(1, 1)      mean, var, skew, kurt = norm.stats(moments = 'mvsk')      x = np.linspace(norm.ppf(0.01),  norm.ppf(0.99), 100)      ax.plot(x, norm.pdf(x),          'r-', lw = 5, alpha = 0.6, label = 'norm pdf')      ax.plot(x, norm.cdf(x),          'b-', lw = 5, alpha = 0.6, label = 'norm cdf')      vals = norm.ppf([0.001, 0.5, 0.999])      np.allclose([0.001, 0.5, 0.999], norm.cdf(vals))      r = norm.rvs(size = 1000)      ax.hist(r, normed = True, histtype = 'stepfilled', alpha = 0.2)      ax.legend(loc = 'best', frameon = False)      mpl.show()  normalDist() Output: Bernoulli Distribution: It is the simplest type of probability distribution. It is a particular case of Binomial distribution, where n=1. It means a binomial distribution takes 'n' number of trials, where n > 1 whereas, the Bernoulli distribution takes only a single trial.   Probability Mass Function of a Bernoulli’s Distribution is:  where p = probability of success and q = probability of failureHere is a code example showing the use of Bernoulli Distribution: from scipy.stats import bernoulli  import seaborn as sb    def bernoulliDist():      data_bern = bernoulli.rvs(size=1200, p = 0.7)      ax = sb.distplot(          data_bern,           kde = True,           color = 'g',           hist_kws = {'alpha' : 1},          kde_kws = {'color': 'y', 'lw': 3, 'label': 'KDE'})      ax.set(xlabel = 'Bernouli Values', ylabel = 'Frequency Distribution')  bernoulliDist() Output:Continuous Uniform Distribution: In this type of continuous distribution, all outcomes are equally possible; each variable gets the same probability of hit as a consequence. This symmetric probabilistic distribution has random variables at an equal interval, with the probability of 1/(b-a). Here is a code example showing the use of Uniform Distribution: from numpy import random  import matplotlib.pyplot as mpl  import seaborn as sb  def uniformDist():      sb.distplot(random.uniform(size = 1200), hist = True)      mpl.show()  uniformDist() Output: Log-Normal Distribution: A Log-Normal distribution is another type of continuous distribution of logarithmic values that form a normal distribution. We can transform a log-normal distribution into a normal distribution. Here is a code example showing the use of Log-Normal Distribution import matplotlib.pyplot as mpl  def lognormalDist():      muu, sig = 3, 1      s = np.random.lognormal(muu, sig, 1000)      cnt, bins, ignored = mpl.hist(s, 80, normed = True, align ='mid', color = 'y')      x = np.linspace(min(bins), max(bins), 10000)      calc = (np.exp( -(np.log(x) - muu) **2 / (2 * sig**2))             / (x * sig * np.sqrt(2 * np.pi)))      mpl.plot(x, calc, linewidth = 2.5, color = 'g')      mpl.axis('tight')      mpl.show()  lognormalDist() Output: Pareto Distribution: It is one of the most critical types of continuous distribution. The Pareto Distribution is a skewed statistical distribution that uses power-law to describe quality control, scientific, social, geophysical, actuarial, and many other types of observable phenomena. The distribution shows slow or heavy-decaying tails in the plot, where much of the data reside at its extreme end. Here is a code example showing the use of Pareto Distribution – import numpy as np  from matplotlib import pyplot as plt  from scipy.stats import pareto  def paretoDist():      xm = 1.5        alp = [2, 4, 6]       x = np.linspace(0, 4, 800)      output = np.array([pareto.pdf(x, scale = xm, b = a) for a in alp])      plt.plot(x, output.T)      plt.show()  paretoDist() Output:Exponential Distribution: It is a type of continuous distribution that determines the time elapsed between events (in a Poisson process). Let’s suppose, that you have the Poisson distribution model that holds the number of events happening in a given period. We can model the time between each birth using an exponential distribution.Here is a code example showing the use of Pareto Distribution – from numpy import random  import matplotlib.pyplot as mpl  import seaborn as sb  def expDist():      sb.distplot(random.exponential(size = 1200), hist = True)      mpl.show()   expDist()Output:Types of the Discrete probability distribution – There are various types of Discrete Probability Distribution a Data science aspirant should know about. Some of them are – Binomial Distribution: It is one of the popular discrete distributions that determine the probability of x success in the 'n' trial. We can use Binomial distribution in situations where we want to extract the probability of SUCCESS or FAILURE from an experiment or survey which went through multiple repetitions. A Binomial distribution holds a fixed number of trials. Also, a binomial event should be independent, and the probability of obtaining failure or success should remain the same. Here is a code example showing the use of Binomial Distribution – from numpy import random  import matplotlib.pyplot as mpl  import seaborn as sb    def binomialDist():      sb.distplot(random.normal(loc = 50, scale = 6, size = 1200), hist = False, label = 'normal')      sb.distplot(random.binomial(n = 100, p = 0.6, size = 1200), hist = False, label = 'binomial')      plt.show()    binomialDist() Output:Geometric Distribution: The geometric probability distribution is one of the crucial types of continuous distributions that determine the probability of any event having likelihood ‘p’ and will happen (occur) after 'n' number of Bernoulli trials. Here 'n' is a discrete random variable. In this distribution, the experiment goes on until we encounter either a success or a failure. The experiment does not depend on the number of trials. Here is a code example showing the use of Geometric Distribution – import matplotlib.pyplot as mpl  def probability_to_occur_at(attempt, probability):      return (1-p)**(attempt - 1) * probability  p = 0.3  attempt = 4  attempts_to_show = range(21)[1:]  print('Possibility that this event will occur on the 7th try: ', probability_to_occur_at(attempt, p))  mpl.xlabel('Number of Trials')  mpl.ylabel('Probability of the Event')  barlist = mpl.bar(attempts_to_show, height=[probability_to_occur_at(x, p) for x in attempts_to_show], tick_label=attempts_to_show)  barlist[attempt].set_color('g')  mpl.show() Output:Poisson Distribution: Poisson distribution is one of the popular types of discrete distribution that shows how many times an event has the possibility of occurrence in a specific set of time. We can obtain this by limiting the Bernoulli distribution from 0 to infinity. Data analysts often use the Poisson distributions to comprehend independent events occurring at a steady rate in a given time interval. Here is a code example showing the use of Poisson Distribution from scipy.stats import poisson  import seaborn as sb  import numpy as np  import matplotlib.pyplot as mpl  def poissonDist():       mpl.figure(figsize = (10, 10))      data_binom = poisson.rvs(mu = 3, size = 5000)      ax = sb.distplot(data_binom, kde=True, color = 'g',                       bins=np.arange(data_binom.min(), data_binom.max() + 1),                       kde_kws={'color': 'y', 'lw': 4, 'label': 'KDE'})      ax.set(xlabel = 'Poisson Distribution', ylabel='Data Frequency')      mpl.show()      poissonDist() Output:Multinomial Distribution: A multinomial distribution is another popular type of discrete probability distribution that calculates the outcome of an event having two or more variables. The term multi means more than one. The Binomial distribution is a particular type of multinomial distribution with two possible outcomes - true/false or heads/tails. Here is a code example showing the use of Multinomial Distribution – import numpy as np  import matplotlib.pyplot as mpl  np.random.seed(99)   n = 12                      pvalue = [0.3, 0.46, 0.22]     s = []  p = []     for size in np.logspace(2, 3):      outcomes = np.random.multinomial(n, pvalue, size=int(size))        prob = sum((outcomes[:,0] == 7) & (outcomes[:,1] == 2) & (outcomes[:,2] == 3))/len(outcomes)      p.append(prob)      s.append(int(size))  fig1 = mpl.figure()  mpl.plot(s, p, 'o-')  mpl.plot(s, [0.0248]*len(s), '--r')  mpl.grid()  mpl.xlim(xmin = 0)  mpl.xlabel('Number of Events')  mpl.ylabel('Function p(X = K)') Output:Negative Binomial Distribution: It is also a type of discrete probability distribution for random variables having negative binomial events. It is also known as the Pascal distribution, where the random variable tells us the number of repeated trials produced during a specific number of experiments.  Here is a code example showing the use of Negative Binomial Distribution – import matplotlib.pyplot as mpl   import numpy as np   from scipy.stats import nbinom    x = np.linspace(0, 6, 70)   gr, kr = 0.3, 0.7        g = nbinom.ppf(x, gr, kr)   s = nbinom.pmf(x, gr, kr)   mpl.plot(x, g, "*", x, s, "r--") Output: Apart from these mentioned distribution types, various other types of probability distributions exist that data science professionals can use to extract reliable datasets. In the next topic, we will understand some interconnections & relationships between various types of probability distributions. Relationship between various Probability distributions – It is surprising to see that different types of probability distributions are interconnected. In the chart shown below, the dashed line is for limited connections between two families of distribution, whereas the solid lines show the exact relationship between them in terms of transformation, variable, type, etc. Conclusion  Probability distributions are prevalent among data analysts and data science professionals because of their wide usage. Today, companies and enterprises hire data science professionals in many sectors, namely, computer science, health, insurance, engineering, and even social science, where probability distributions appear as fundamental tools for application. It is essential for Data analysts and data scientists. to know the core of statistics. Probability Distributions perform a requisite role in analyzing data and cooking a dataset to train the algorithms efficiently. If you want to learn more about data science - particularly probability distributions and their uses, check out KnowledgeHut's comprehensive Data science course. 
9639
Types of Probability Distributions Every Data Scie...

Data Science has become one of the most popular in... Read More

Role of Unstructured Data in Data Science

Data has become the new game changer for businesses. Typically, data scientists categorize data into three broad divisions - structured, semi-structured, and unstructured data. In this article, you will get to know about unstructured data, sources of unstructured data, unstructured data vs. structured data, the use of structured and unstructured data in machine learning, and the difference between structured and unstructured data. Let us first understand what is unstructured data with examples. What is unstructured data? Unstructured data is a kind of data format where there is no organized form or type of data. Videos, texts, images, document files, audio materials, email contents and more are considered to be unstructured data. It is the most copious form of business data, and cannot be stored in a structured database or relational database. Some examples of unstructured data are the photos we post on social media platforms, the tagging we do, the multimedia files we upload, and the documents we share. Seagate predicts that the global data-sphere will expand to 163 zettabytes by 2025, where most of the data will be in the unstructured format. Characteristics of Unstructured DataUnstructured data cannot be organized in a predefined fashion, and is not a homogenous data model. This makes it difficult to manage. Apart from that, these are the other characteristics of unstructured data. You cannot store unstructured data in the form of rows and columns as we do in a database table. Unstructured data is heterogeneous in structure and does not have any specific data model. The creation of such data does not follow any semantics or habits. Due to the lack of any particular sequence or format, it is difficult to manage. Such data does not have an identifiable structure. Sources of Unstructured Data There are various sources of unstructured data. Some of them are: Content websites Social networking sites Online images Memos Reports and research papers Documents, spreadsheets, and presentations Audio mining, chatbots Surveys Feedback systems Advantages of Unstructured Data Unstructured data has become exceptionally easy to store because of MongoDB, Cassandra, or even using JSON. Modern NoSQL databases and software allows data engineers to collect and extract data from various sources. There are numerous benefits that enterprises and businesses can gain from unstructured data. These are: With the advent of unstructured data, we can store data that lacks a proper format or structure. There is no fixed schema or data structure for storing such data, which gives flexibility in storing data of different genres. Unstructured data is much more portable by nature. Unstructured data is scalable and flexible to store. Database systems like MongoDB, Cassandra, etc., can easily handle the heterogeneous properties of unstructured data. Different applications and platforms produce unstructured data that becomes useful in business intelligence, unstructured data analytics, and various other fields. Unstructured data analysis allows finding comprehensive data stories from data like email contents, website information, social media posts, mobile data, cache files and more. Unstructured data, along with data analytics, helps companies improve customer experience. Detection of the taste of consumers and their choices becomes easy because of unstructured data analysis. Disadvantages of Unstructured data Storing and managing unstructured data is difficult because there is no proper structure or schema. Data indexing is also a substantial challenge and hence becomes unclear due to its disorganized nature. Search results from an unstructured dataset are also not accurate because it does not have predefined attributes. Data security is also a challenge due to the heterogeneous form of data. Problems faced and solutions for storing unstructured data. Until recently, it was challenging to store, evaluate, and manage unstructured data. But with the advent of modern data analysis tools, algorithms, CAS (content addressable storage system), and big data technologies, storage and evaluation became easy. Let us first take a look at the various challenges used for storing unstructured data. Storing unstructured data requires a large amount of space. Indexing of unstructured data is a hectic task. Database operations such as deleting and updating become difficult because of the disorganized nature of the data. Storing and managing video, audio, image file, emails, social media data is also challenging. Unstructured data increases the storage cost. For solving such issues, there are some particular approaches. These are: CAS system helps in storing unstructured data efficiently. We can preserve unstructured data in XML format. Developers can store unstructured data in an RDBMS system supporting BLOB. We can convert unstructured data into flexible formats so that evaluating and storage becomes easy. Let us now understand the differences between unstructured data vs. structured data. Unstructured Data Vs. Structured Data In this section, we will understand the difference between structured and unstructured data with examples. STRUCTUREDUNSTRUCTUREDStructured data resides in an organized format in a typical database.Unstructured data cannot reside in an organized format, and hence we cannot store it in a typical database.We can store structured data in SQL database tables having rows and columns.Storing and managing unstructured data requires specialized databases, along with a variety of business intelligence and analytics applications.It is tough to scale a database schema.It is highly scalable.Structured data gets generated in colleges, universities, banks, companies where people have to deal with names, date of birth, salary, marks and so on.We generate or find unstructured data in social media platforms, emails, analyzed data for business intelligence, call centers, chatbots and so on.Queries in structured data allow complex joining.Unstructured data allows only textual queries.The schema of a structured dataset is less flexible and dependent.An unstructured dataset is flexible but does not have any particular schema.It has various concurrency techniques.It has no concurrency techniques.We can use SQL, MySQL, SQLite, Oracle DB, Teradata to store structured data.We can use NoSQL (Not Only SQL) to store unstructured data.Types of Unstructured Data Do you have any idea just how much of unstructured data we produce and from what sources? Unstructured data includes all those forms of data that we cannot actively manage in an RDBMS system that is a transactional system. We can store structured data in the form of records. But this is not the case with unstructured data. Before the advent of object-based storage, most of the unstructured data was stored in file-based systems. Here are some of the types of unstructured data. Rich media content: Entertainment files, surveillance data, multimedia email attachments, geospatial data, audio files (call center and other recorded audio), weather reports (graphical), etc., comes under this genre. Document data: Invoices, text-file records, email contents, productivity applications, etc., are included under this genre. Internet of Things (IoT) data: Ticker data, sensor data, data from other IoT devices come under this genre. Apart from all these, data from business intelligence and analysis, machine learning datasets, and artificial intelligence data training datasets are also a separate genre of unstructured data. Examples of Unstructured Data There are various sources from where we can obtain unstructured data. The prominent use of this data is in unstructured data analytics. Let us now understand what are some examples of unstructured data and their sources – Healthcare industries generate a massive volume of human as well as machine-generated unstructured data. Human-generated unstructured data could be in the form of patient-doctor or patient-nurse conversations, which are usually recorded in audio or text formats. Unstructured data generated by machines includes emergency video camera footage, surgical robots, data accumulated from medical imaging devices like endoscopes, laparoscopes and more.  Social Media is an intrinsic entity of our daily life. Billions of people come together to join channels, share different thoughts, and exchange information with their loved ones. They create and share such data over social media platforms in the form of images, video clips, audio messages, tagging people (this helps companies to map relations between two or more people), entertainment data, educational data, geolocations, texts, etc. Other spectra of data generated from social media platforms are behavior patterns, perceptions, influencers, trends, news, and events. Business and corporate documents generate a multitude of unstructured data such as emails, presentations, reports containing texts, images, presentation reports, video contents, feedback and much more. These documents help to create knowledge repositories within an organization to make better implicit operations. Live chat, video conferencing, web meeting, chatbot-customer messages, surveillance data are other prominent examples of unstructured data that companies can cultivate to get more insights into the details of a person. Some prominent examples of unstructured data used in enterprises and organizations are: Reports and documents, like Word files or PDF files Multimedia files, such as audio, images, designed texts, themes, and videos System logs Medical images Flat files Scanned documents (which are images that hold numbers and text – for example, OCR) Biometric data Unstructured Data Analytics Tools  You might be wondering what tools can come into use to gather and analyze information that does not have a predefined structure or model. Various tools and programming languages use structured and unstructured data for machine learning and data analysis. These are: Tableau MonkeyLearn Apache Spark SAS Python MS. Excel RapidMiner KNIME QlikView Python programming R programming Many cloud services (like Amazon AWS, Microsoft Azure, IBM Cloud, Google Cloud) also offer unstructured data analysis solutions bundled with their services. How to analyze unstructured data? In the past, the process of storage and analysis of unstructured data was not well defined. Enterprises used to carry out this kind of analysis manually. But with the advent of modern tools and programming languages, most of the unstructured data analysis methods became highly advanced. AI-powered tools use algorithms designed precisely to help to break down unstructured data for analysis. Unstructured data analytics tools, along with Natural language processing (NLP) and machine learning algorithms, help advanced software tools analyze and extract analytical data from the unstructured datasets. Before using these tools for analyzing unstructured data, you must properly go through a few steps and keep these points in mind. Set a clear goal for analyzing the data: It is essential to clear your intention about what insights you want to extract from your unstructured data. Knowing this will help you distinguish what type of data you are planning to accumulate. Collect relevant data: Unstructured data is available everywhere, whether it's a social media platform, online feedback or reviews, or a survey form. Depending on the previous point, that is your goal - you have to be precise about what data you want to collect in real-time. Also, keep in mind whether your collected details are relevant or not. Clean your data: Data cleaning or data cleansing is a significant process to detect corrupt or irrelevant data from the dataset, followed by modifying or deleting the coarse and sloppy data. This phase is also known as the data-preprocessing phase, where you have to reduce the noise, carry out data slicing for meaningful representation, and remove unnecessary data. Use Technology and tools: Once you perform the data cleaning, it is time to utilize unstructured data analysis tools to prepare and cultivate the insights from your data. Technologies used for unstructured data storage (NoSQL) can help in managing your flow of data. Other tools and programming libraries like Tableau, Matplotlib, Pandas, and Google Data Studio allows us to extract and visualize unstructured data. Data can be visualized and presented in the form of compelling graphs, plots, and charts. How to Extract information from Unstructured Data? With the growth in digitization during the information era, repetitious transactions in data cause data flooding. The exponential accretion in the speed of digital data creation has brought a whole new domain of understanding user interaction with the online world. According to Gartner, 80% of the data created by an organization or its application is unstructured. While extracting exact information through appropriate analysis of organized data is not yet possible, even obtaining a decent sense of this unstructured data is quite tough. Until now, there are no perfect tools to analyze unstructured data. But algorithms and tools designed using machine learning, Natural language processing, Deep learning, and Graph Analysis (a mathematical method for estimating graph structures) help us to get the upper hand in extracting information from unstructured data. Other neural network models like modern linguistic models follow unsupervised learning techniques to gain a good 'knowledge' about the unstructured dataset before going into a specific supervised learning step. AI-based algorithms and technologies are capable enough to extract keywords, locations, phone numbers, analyze image meaning (through digital image processing). We can then understand what to evaluate and identify information that is essential to your business. ConclusionUnstructured data is found abundantly from sources like documents, records, emails, social media posts, feedbacks, call-records, log-in session data, video, audio, and images. Manually analyzing unstructured data is very time-consuming and can be very boring at the same time. With the growth of data science and machine learning algorithms and models, it has become easy to gather and analyze insights from unstructured information.  According to some research, data analytics tools like MonkeyLearn Studio, Tableau, RapidMiner help analyze unstructured data 1200x faster than the manual approach. Analyzing such data will help you learn more about your customers as well as competitors. Text analysis software, along with machine learning models, will help you dig deep into such datasets and make you gain an in-depth understanding of the overall scenario with fine-grained analyses.
5797
Role of Unstructured Data in Data Science

Data has become the new game changer for busines... Read More

What Is Statistical Analysis and Its Business Applications?

Statistics is a science concerned with collection, analysis, interpretation, and presentation of data. In Statistics, we generally want to study a population. You may consider a population as a collection of things, persons, or objects under experiment or study. It is usually not possible to gain access to all of the information from the entire population due to logistical reasons. So, when we want to study a population, we generally select a sample. In sampling, we select a portion (or subset) of the larger population and then study the portion (or the sample) to learn about the population. Data is the result of sampling from a population.Major ClassificationThere are two basic branches of Statistics – Descriptive and Inferential statistics. Let us understand the two branches in brief. Descriptive statistics Descriptive statistics involves organizing and summarizing the data for better and easier understanding. Unlike Inferential statistics, Descriptive statistics seeks to describe the data, however, it does not attempt to draw inferences from the sample to the whole population. We simply describe the data in a sample. It is not developed on the basis of probability unlike Inferential statistics. Descriptive statistics is further broken into two categories – Measure of Central Tendency and Measures of Variability. Inferential statisticsInferential statistics is the method of estimating the population parameter based on the sample information. It applies dimensions from sample groups in an experiment to contrast the conduct group and make overviews on the large population sample. Please note that the inferential statistics are effective and valuable only when examining each member of the group is difficult. Let us understand Descriptive and Inferential statistics with the help of an example. Task – Suppose, you need to calculate the score of the players who scored a century in a cricket tournament.  Solution: Using Descriptive statistics you can get the desired results.   Task – Now, you need the overall score of the players who scored a century in the cricket tournament.  Solution: Applying the knowledge of Inferential statistics will help you in getting your desired results.  Top Five Considerations for Statistical Data AnalysisData can be messy. Even a small blunder may cost you a fortune. Therefore, special care when working with statistical data is of utmost importance. Here are a few key takeaways you must consider to minimize errors and improve accuracy. Define the purpose and determine the location where the publication will take place.  Understand the assets to undertake the investigation. Understand the individual capability of appropriately managing and understanding the analysis.  Determine whether there is a need to repeat the process.  Know the expectation of the individuals evaluating reviewing, committee, and supervision. Statistics and ParametersDetermining the sample size requires understanding statistics and parameters. The two being very closely related are often confused and sometimes hard to distinguish.  StatisticsA statistic is merely a portion of a target sample. It refers to the measure of the values calculated from the population.  A parameter is a fixed and unknown numerical value used for describing the entire population. The most commonly used parameters are: Mean Median Mode Mean :  The mean is the average or the most common value in a data sample or a population. It is also referred to as the expected value. Formula: Sum of the total number of observations/the number of observations. Experimental data set: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20  Calculating mean:   (2 + 4 + 6 + 8 + 10 + 12 + 14 + 16 + 18 + 20)/10  = 110/10   = 11 Median:  In statistics, the median is the value separating the higher half from the lower half of a data sample, a population, or a probability distribution. It’s the mid-value obtained by arranging the data in increasing order or descending order. Formula:  Let n be the data set (increasing order) When data set is odd: Median = n+1/2th term Case-I: (n is odd)  Experimental data set = 1, 2, 3, 4, 5  Median (n = 5) = [(5 +1)/2]th term      = 6/2 term       = 3rd term   Therefore, the median is 3 When data set is even: Median = [n/2th + (n/2 + 1)th] /2 Case-II: (n is even)  Experimental data set = 1, 2, 3, 4, 5, 6   Median (n = 6) = [n/2th + (n/2 + 1)th]/2     = ( 6/2th + (6/2 +1)th]/2     = (3rd + 4th)/2      = (3 + 4)/2      = 7/2      = 3.5  Therefore, the median is 3.5 Mode: The mode is the value that appears most often in a set of data or a population. Experimental data set= 1, 2, 2, 2, 3, 3, 3, 3, 3, 4, 4,4,5, 6  Mode = 3 (Since 3 is the most repeated element in the sequence.) Terms Used to Describe DataWhen working with data, you will need to search, inspect, and characterize them. To understand the data in a tech-savvy and straightforward way, we use a few statistical terms to denote them individually or in groups.  The most frequently used terms used to describe data include data point, quantitative variables, indicator, statistic, time-series data, variable, data aggregation, time series, dataset, and database. Let us define each one of them in brief: Data points: These are the numerical files formed and organized for interpretations. Quantitative variables: These variables present the information in digit form.  Indicator: An indicator explains the action of a community's social-economic surroundings.  Time-series data: The time-series defines the sequential data.  Data aggregation: A group of data points and data set. Database: A group of arranged information for examination and recovery.  Time-series: A set of measures of a variable documented over a specified time. Step-by-Step Statistical Analysis ProcessThe statistical analysis process involves five steps followed one after another. Step 1: Design the study and find the population of the study. Step 2: Collect data as samples. Step 3: Describe the data in the sample. Step 4: Make inferences with the help of samples and calculations Step 5: Take action Data distributionData distribution is an entry that displays entire imaginable readings of data. It shows how frequently a value occurs. Distributed data is always in ascending order, charts, and graphs enabling visibility of measurements and frequencies. The distribution function displaying the density of values of reading is known as the probability density function. Percentiles in data distributionA percentile is the reading in a distribution with a specified percentage of clarifications under it.  Let us understand percentiles with the help of an example.  Suppose you have scored 90th percentile on a math test. A basic interpretation is that merely 4-5% of the scores were higher than your scores. Right? The median is 50th percentile because the assumed 50% of the values are higher than the median. Dispersion Dispersion explains the magnitude of distribution readings anticipated for a specific variable and multiple unique statistics like range, variance, and standard deviation. For instance, high values of a data set are widely scattered while small values of data are firmly clustered. Histogram The histogram is a pictorial display that arranges a group of data facts into user detailed ranges. A histogram summarizes a data series into a simple interpreted graphic by obtaining many data facts and combining them into reasonable ranges. It contains a variety of results into columns on the x-axis. The y axis displays percentages of data for each column and is applied to picture data distributions. Bell Curve distribution Bell curve distribution is a pictorial representation of a probability distribution whose fundamental standard deviation obtained from the mean makes the bell, shaped curving. The peak point on the curve symbolizes the maximum likely occasion in a pattern of data. The other possible outcomes are symmetrically dispersed around the mean, making a descending sloping curve on both sides of the peak. The curve breadth is therefore known as the standard deviation. Hypothesis testingHypothesis testing is a process where experts experiment with a theory of a population parameter. It aims to evaluate the credibility of a hypothesis using sample data. The five steps involved in hypothesis testing are:  Identify the no outcome hypothesis.  (A worthless or a no-output hypothesis has no outcome, connection, or dissimilarities amongst many factors.) Identify the alternative hypothesis.  Establish the importance level of the hypothesis.  Estimate the experiment statistic and equivalent P-value. P-value explains the possibility of getting a sample statistic.  Sketch a conclusion to interpret into a report about the alternate hypothesis. Types of variablesA variable is any digit, amount, or feature that is countable or measurable. Simply put, it is a variable characteristic that varies. The six types of variables include the following: Dependent variableA dependent variable has values that vary according to the value of another variable known as the independent variable.  Independent variableAn independent variable on the other side is controllable by experts. Its reports are recorded and equated.  Intervening variableAn intervening variable explicates fundamental relations between variables. Moderator variableA moderator variable upsets the power of the connection between dependent and independent variables.  Control variableA control variable is anything restricted to a research study. The values are constant throughout the experiment. Extraneous variableExtraneous variable refers to the entire variables that are dependent but can upset experimental outcomes. Chi-square testChi-square test records the contrast of a model to actual experimental data. Data is unsystematic, underdone, equally limited, obtained from independent variables, and a sufficient sample. It relates the size of any inconsistencies among the expected outcomes and the actual outcomes, provided with the sample size and the number of variables in the connection. Types of FrequenciesFrequency refers to the number of repetitions of reading in an experiment in a given time. Three types of frequency distribution include the following: Grouped, ungrouped Cumulative, relative Relative cumulative frequency distribution. Features of FrequenciesThe calculation of central tendency and position (median, mean, and mode). The measure of dispersion (range, variance, and standard deviation). Degree of symmetry (skewness). Peakedness (kurtosis). Correlation MatrixThe correlation matrix is a table that shows the correlation coefficients of unique variables. It is a powerful tool that summarises datasets points and picture sequences in the provided data. A correlation matrix includes rows and columns that display variables. Additionally, the correlation matrix exploits in aggregation with other varieties of statistical analysis. Inferential StatisticsInferential statistics use random data samples for demonstration and to create inferences. They are measured when analysis of each individual of a whole group is not likely to happen. Applications of Inferential StatisticsInferential statistics in educational research is not likely to sample the entire population that has summaries. For instance, the aim of an investigation study may be to obtain whether a new method of learning mathematics develops mathematical accomplishment for all students in a class. Marketing organizations: Marketing organizations use inferential statistics to dispute a survey and request inquiries. It is because carrying out surveys for all the individuals about merchandise is not likely. Finance departments: Financial departments apply inferential statistics for expected financial plan and resources expenses, especially when there are several indefinite aspects. However, economists cannot estimate all that use possibility. Economic planning: In economic planning, there are potent methods like index figures, time series investigation, and estimation. Inferential statistics measures national income and its components. It gathers info about revenue, investment, saving, and spending to establish links among them. Key TakeawaysStatistical analysis is the gathering and explanation of data to expose sequences and tendencies.   Two divisions of statistical analysis are statistical and non-statistical analyses.  Descriptive and Inferential statistics are the two main categories of statistical analysis. Descriptive statistics describe data, whereas Inferential statistics equate dissimilarities between the sample groups.  Statistics aims to teach individuals how to use restricted samples to generate intellectual and precise results for a large group.   Mean, median, and mode are the statistical analysis parameters used to measure central tendency.   Conclusion Statistical analysis is the procedure of gathering and examining data to recognize sequences and trends. It uses random samples of data obtained from a population to demonstrate and create inferences on a group. Inferential statistics applies economic planning with potent methods like index figures, time series investigation, and estimation.  Statistical analysis finds its applications in all the major sectors – marketing, finance, economic, operations, and data mining. Statistical analysis aids marketing organizations in disputing a survey and requesting inquiries concerning their merchandise. 
5886
What Is Statistical Analysis and Its Business Appl...

Statistics is a science concerned with collection,... Read More