Ordinary Least Squares Regression Model

Project description
Suppose you are going to use Ordinary Least Squares regression to build a predictive model. You have 1 MM data records that consists of a response variable Y and 100 possible predictor variables. How should you approach the modeling process?

In particular:

(1) How should you start the modeling process?

(2) How many records should you use in the modeling process?

(3) How will you select your predictor variables? We will not want to use all 100 predictor variables in our model.

(4) How will you assess the predictive accuracy of your model? Are there any particular metrics? Should you use a statistical approach such as cross-validation? Why would we want to consider the use of cross-validation?
FOR MORE INFORMATION ON THIS TOPIC CLICK HERE.