Fuzzy K-Means Incremental Clustering Based on K-Center and Vector Quantization

Fuzzy k-means and vector quantization are combined in this paper to complement each other in incremental mode because each has qualities which the other lacks. The threshold of vector quantization is given and the pattern of computing the distance between the new coming data point and the k centers is introduced in a new way. We firstly reduce redundant attributes and eliminate the difference of units of dimensions and make units of all attributes same. Then, we use k-center to produce initial k means and partition data points into no more than k clusters. Besides, we adopt vector quantization to classify incremental data points and then adjust means after the structure of clustering varying. Finally, it is applied to real datasets and results show its efficiency and precision.