There are various methods to add or remove variables to determine the best possible model.
In the backward method, iterations start with considering all the variables and we will remove variables one by one until all the prescribed statistics are met (such as no insignificance and multi-collinearity, and so on). Finally, the overall statistic will be checked, such as if R-squared value is > 0.7 , it is considered a good model, else reject it. In industry, practitioners mainly prefer to work on backward methods.
In the case of forward, we will start with no variables and keep on adding significant variables until the overall model's fit improves.
In the following method, we have used the backward selection method, ...
