statsmodels|sklearn|matplotlib|pandas|numpy

ols|summary()

Task : build a multilinear regression model to predict profit

Summary : I build a multilinear regression model using : ols method various steps for multilinear regression models are :

1.EDA

• columns names
• columns data types
• duplicacy in records
• outlier detection
• etc...

2.checking linearity

• pairplot

3.correlation analysis

• correlation (relation between x and y) ++ (point to notice - 1 : is there any relation between x and x)
• building model (trial mode) ++ (point to notice - 2 : rquared, AIC, p-value, adj_rsquared)

5.dealing with problematic columns

• a-1) checking if x and x have good correlation
• a-2) is p-value significant or not
• multicollinearity (VIF) (checking how individual feature affecting the model prediction)
• error / residue handing
• improving model (trial mode)

6.dealing with problematic rows

• checking influential points (cook's distance)
• improving model (trial mode)

7. final model

After dealing with all the above case we build our final model. Final model has good R-squared value, less AIC value, significal p-value. Having all these requirements, we can now approve our model.

Conclusion :

[model1 - 0.961758 (R-squared)] [model2 - 0.961316 (R-squared)] [model3 - 0.961086 (R-squared)] [final_model - 0.965359 (R-squared)]