Summary

In this chapter, the goal was to use a small dataset to provide an introduction to practically apply an advanced feature selection for linear models. The outcome for our data was quantitative but the glmnet package that we used will also support qualitative outcomes (binomial and multinomial classifications). An introduction to regularization and the three techniques that incorporate it were provided and utilized to build and compare models. Regularization is a powerful technique to improve computational efficiency and to possibly extract more meaningful features versus the other modeling techniques. Additionally, we started to use the caret package to optimize multiple parameters when training a model. Up to this point, we've been purely talking about linear models. In the next couple of chapters, we will begin to use nonlinear models for both classification and regression problems.