Abstract:

An improved affinity propagation algorithm based on coefficient of variation was proposed via feature weighing of samples, which has overcome the impact of redundant information, and the new similarity measure method was proposed. The experimental results show that the proposed algorithm is not only quick and efficient but also better than the traditional affinity propagation algorithm and the classical Kmeans method for clustering when it was used to process the samples with more characteristics and attributes, and information overlap.

Key words: affinity propagation algorithm, coefficient of variation, feature weighing, clustering