关键词: 聚类, 粒计算, 粗糙集, 属性约简

Abstract: Against such shortcomings of the traditional fish-swarm algorithm with the effect of initial fish-group random, easily falling into a local extremum, low efficiency of handling boundary data and low clustering accuracy, an improved artificial fish swarm algorithm based on Rough set and granular computing is proposed. Initially, the algorithm introduces the granular computing theory and initializes the fish group by the density and max-min distance means so that the algorithm avoids being effected by random. Meantime the algorithm is combined with rough set and attribute reduction and decision system to resolve the clustering problem of boundary data. With the principles of within-class compactness and between-class separability while designing the fitness function, it also can be regarded as the termination condition of algorithm. Experiment results show that the algorithm has enhanced the accuracy and the abilities to obtain global extremum and embodies better clustering performance.

Key words: clustering, granular computing, rough set, attribute reduction