Abstract: Aiming at the problem that the traditional fuzzy clustering algorithm needed to set the parameters and the initial clustering centers in advance, which resulted in unstable clustering results, we proposed a dynamic fuzzy clustering algorithm based on weighted difference degree. Firstly, the concepts of the sample feature weight vector and the difference degree between s
amples were introduced to describe the distribution of datasets, and a new evaluation index was used to obtain candidate clustering centers. Secondly, according to the criterion of minimum difference degree, the remaining sample points were classified. Finally, the candidate clustering centers were selected and merged according to the evaluation criteria of the DaviesBouldin index (DBI). The experimental results show that the performance of this algorithm is significantly better than that of traditional clustering algorithmin on different datasets, and it has higher adaptability and stability.

Key words: fuzzy clustering algorithm, weight vector, difference degree, Davies-Bouldin index (DBI), self-adaptive