An Enhanced K-Nearest Neighbor Algorithm Using Information Gain and Clustering

KNN (k-nearest neighbor) is an extensively used classification algorithm owing to its simplicity, ease of implementation and effectiveness. It is one of the top ten data mining algorithms, has been widely applied in various fields. KNN has few shortcomings affecting its accuracy of classification. It has large memory requirements as well as high time complexity. Several techniques have been proposed to improve these shortcomings in literature. In this paper, we have first reviewed some improvements made in KNN algorithm. Then, we have proposed our novel improved algorithm. It is a combination of dynamic selected, attribute weighted and distance weighted techniques. We have experimentally tested our proposed algorithm in Net Beans IDE, using a standard UCI dataset-Iris. The accuracy of our algorithm is improved with a blend of classification and clustering techniques. Experimental results have proved that our proposed algorithm performs better than conventional KNN algorithm.