Feature Weight and Its Application in Weight Determination of Medical Scale Items

Abstract

Actually, the determination of medical scales items is feature weight problem in data-mining area. The framework of EC-based (Evolutionary computation) classification method for feature weight is presented contrasted with traditional statistical methods. And an improved EC-based k-NN algorithm for feature weight, GS-k-NN, is put forward and presented. Comparison between PSO and GA is made as well as among k-NN, GS-k-NN, C4.5, SVM in the paper. Results show that PSO-based GS-k-NN is more effective than other algorithms.