Feature selection is the process of identifying and retaining only the most relevant features for use in model training, and removing irrelevant or redundant ones. The goal is to simplify the model, reduce overfitting, and improve generalization by eliminating noise or collinear features. Feature selection methods can be categorized as filter methods (select features by ranking via statistical tests or information measures, independent of any model – e.g., chi-squared test, mutual information), wrapper methods (select features by training models and using predictive performance as feedback – e.g., recursive feature elimination), and embedded methods (feature selection happens as part of the model training – e.g., Lasso regularization drives some coefficients to zero, effectively selecting features). By selecting a subset, we also reduce computational cost. However, one must be careful to perform feature selection considering validation to ensure selected features truly help generalization.