Feature Selection is a process where we select a subset of features which contribute most to the output variable. If there are n features and out of them only few features have relavance with the output variable then, instead of training the model with all the n features we can select only few features so the performance of the model increases and could overcome the problem of curse of dimensionality.
-
Filter Method
checks the relevance of features with output variable.
1) CHI squared test 2) ANOVA test 3) Correlation Coefficient
-
Wrapper methods
1) forward Selection 2) Backward elimination
-
Embedded methods
Learns the feature selection while building the model.
1) Decision Tree