Effect of Principal Component Analysis on Genetic Algorithm Feature Selection

A genetic algorithm feature selection module is used to find the optimal number of features that yield the best quality classifiers. Two groups of data are used and compared. The first group uses original data features in creating population. In the second group, the principal component analysis technique is used on data sets from the first group to obtain principal components. These principal components are used to create population instead of original features. The results indicated that genetic algorithm works well with both sets of data but did not show any improvement in terms of execution time or accuracy score. However, the interpretation of the results from the genetic algorithm feature selection module in the second case is easier to understand. Since principal components are sorted in descending order, it can be clearly seen that the components with higher significance tend to always be chosen by the genetic algorithm.