Abstract:

The basic idea for the K-means algorithm is to partition all elements to different clusters by iterative method so that the elements in the same cluster have the minimum dissimilarity and the elements in different clusters have the maximum dissimilarity. However, it may simply cluster these elements in an overlap which should be in different clusters to the same cluster, so the result of clustering cannot show the element's overlap characteristic. In the paper, a new Kmeans Algorithm based by fuzzy logic is proposed. It can not only obtain the same clustering result as original algorithm but also get element's overlap characteristic. Experiment shows that the new algorithm is efficiency.

Key words: fuzzy logic;K-means algorithm;overlap characteristic