6.2: Different Feature Selection Methods
There are three broad approaches to
feature selection. Filter method, Wrapper method, and Embedded method. We will
discuss the first two methods in this chapter. For certain modeling techniques,
feature selection and model training happen in a parallel fashion. These are
called embedded methods of feature selection. Also, for certain modeling
techniques, feature selection should be approached after keeping in mind the
nuances of the technique. For certain techniques, feature selection is
altogether discouraged. We will discuss these fine nuances in detail in chapter
7.
Apart from filter, wrapper, and
embedded methods, there is another, 4th group of feature selection techniques.
These are known as metaheuristic algorithms. In chapter 8, we will discuss
these techniques in detail.