Investigating a novel GA-based feature selection method using improved KNN classifiers
Amit Kumar Saxena, Dovendra Patre, Abhishek Kumar Dubey · International Journal of Information and Communication Technology · 2011
This paper presents a novel scheme to select a subset of features from a dataset. We apply genetic algorithm (GA) with a random small subset of features. The GA explores stochastically a better subset of features using various combinations of lengths and features over a number of generations. The classification accuracy due to different classifiers in presence of these subsets of features is taken as the performance criteria (objective function) of GA. The proposed scheme is tested on a few UCI datasets. The performances of the KNN, informative KNN (local LI-KNN and global GI-KNN), and LI-KNN with boosting in presence of all features and those in presence of only selected subset of features are compared with reported results. With extensive simulation study, it is observed that the proposed scheme produces a reasonably good accuracy with a reduced subset of features in these datasets.