earticle

영어

K-means algorithm is sensitive to initial cluster centers and its solutions are apt to be trapped in local optimums. In order to solve these problems, we propose an optimized ar-tificial bee colony algorithm for clustering. The proposed method first obtains optimized sources by improving the selection of the initial clustering centers; then, uses a novel dy-namic local optimization strategy utilizing roulette wheel selection algorithm for further enhancing local optimization. To prove its effectiveness, we validate the proposed algo-rithm on four datasets from UCI and compared the results with K-means, K-means++ and Artificial Bee Colony algorithm. Experiment results show that the proposed algo-rithm performs better than other clustering algorithms.