10 Feature Selection Overview
Chapters 5 through 9 have provided tools for engineering features (or predictors) to put them in a form that enables models to better find the predictive signal relative to the outcome. Some of these techniques, like 1-1 transformations or dimensions reduction methods, lead to a new predictor set that has the as many or fewer predictors than the original data. Other transformations, like basis expansions, generate more features than the original data. The hope is that some of the newly engineered predictors capture a predictive relationship with the outcome. But some may not be relevant to the outcome. Moreover, many of the original predictors also may not contain predictive information. For a number of models, predictive performance is degraded as the number of uninformative predictors increases. Therefore, there is a genuine need to appropriately select predictors for modeling.
The remaining chapters will focus on supervised feature selection where the choice of which predictors to retain is guided by their affect on the outcome. This chapter will provide an introduction to feature selection, the general nomenclature for these methods, and some notable pitfalls.