K value of k-means algorithm based on granular computing and its application
Bian Cai-fen · Jisuanji gongcheng yu sheji · 2015
To reduce effects of k value selection of clustering analysis,an improvement of clustering validity function was used to select k value.Attributed clustering validity function considered not only how to avoid the effect caused by data but also the difference and similarity between classes comprehensively.By using attribute resolution,the influence on the similarity within classes and the difference between classes caused by large or small attribute value was avoided.The correctness of the improved clustering validity function was verified by some datasets of UCI database and k-means algorithm.The improved clustering validity was used in corn breeding.Experimental results show that the improved clustering validity function is correct and effective in the corn samples selection while breeding.