Improving the usability of gaknn framework

Abstract

K nearest neighbour classification (KNN) is a popular non parametric and lazy algorithm for classification. gaKnn framework is a implementation of the KNN algorithm combine with genetic algorithm. It provides genetic algorithm optimization for KNN algorithm which will optimize the weight values for each attribute and k value. In this paper, I proposed improvements for the current implementation of the gaKnn framework to improve its usability and performance using kd tree to improve the KNN algorithm, different data and file type usage and regression algorithm based on k nearest neighbour. Mainly it introduce three modules for the current implementation of the gaKnn framework namely csv file reader and writer module, large dataset module and KNN regression module.