When Target Variable is Numerical:
Here the target variable is numerical. Previously we the target variable was categorical(both binary and multi class). Let us use the dataset obtained from Kaggle’s “Student Performance (Multiple Linear Regression)” dataset. This dataset contains one target variable (PerformIndex) and Hours studied, Previous scores, Extracurricular Activities, Sleeping Hours, Same Sample questions Practiced(5) independent variables. for our convenience sake we have changed the label of features and target
The performance index represents the student’s academic performance and has been rounded to the nearest integer.
The index ranges from 10 to 100, with higher values indicating better performance. The data contains 10000 records
Load libraries and read data
Information about data
Statistical Information
Check if any null value exist?
Separate features(X) and target(y)
DecisionTreeRegressor: criterion-Mean Squared_error(mse)
Tree using Squared_error
Decision Tree using default-Squared_error criterion + mean value
Node statistics
show the results after converting the array into DF
Result shows only the leave-less children nodes and their respective values
DecisionTreeRegressor: criterion-Mean Absolute_error(mae)
Decision Tree using Absolute-error Criterion
Tree Impurity
DecisionTreeRegressor: criterion-Friedman mse
Decision Tree using Friedman mse Criterion
Tree Impurity
Node Statistics
show the results after converting the array into DF
DecisionTreeRegressor: criterion-Poisson
Decision Tree using Poisson Criterion
Node Statistics
Make X and y as an array
Using Statsmodels’ find regression under OLS method
For the above we have already imported statsmodels.api as sm
Summary
The formula for predicting Performance Index
yhat = -34.0756 +2.8530*Hours+1.0184*Previscores+0.6129*Extraactivities+0.4806*Sleepinghrs+0.1938*Practisedsamequest
Correlation
PreviousScores correlation is very high nearer to 1. So, if influences Performance Index in a more significant way than other features. the next high one is meant for Hours
Feature selection Using Correlation
Feature selection Using CHI-SQUARE
Selected features based on Chi-Square
Feature Selection using Univariate, SelectKBest and chi2
Accuracy Using SKlearn
