Summary

In this chapter, the goal was to use a simulated dataset to provide an introduction to learning how to apply advanced feature selection for linear and generalized linear models. We used the glmnet package to predict class probabilities for a binary classification problem using logistic regression. These methods can be adapted to linear regression and multinomial classifications. An introduction to regularization and the three techniques that incorporate it was provided and utilized to build and compare models. Regularization is a powerful technique to improve computational efficiency and to possibly extract more meaningful features when compared to the other modeling techniques. We saw how to use various performance metrics to compare and select the most appropriate model. 

Up to this point, we've been purely talking about linear and generalized linear models. In the next couple of chapters, we'll begin to use more complex nonlinear models for both classification and regression problems we'll encounter in further chapters.