Research on Attribute-weighting Cluster Algorithm
Shao-Hua Mi · Journal of Chinese Computer Systems · 2012
In an actual cluster problem,the contribution of each attribute is always not the same,the important attributes have priorities,however,each attribute is given the same weight in the traditional cluster algorithm.It is can improve the effect of cluster if the important attributes are granted the comparatively large weights.In this paper,Modified Particle Swarm Optimization algorithm(MPSO) is put forward to get the attribute weights for the features for a cluster problem.Then these attribute weights are brought into Iterative Self-Organizing Data Analysis Techniques Algorithm(ISODATA) to construct attribute-weighting ISODATA algorithm based on MPSO algorithm(MPSO-WISODATA).Extensive experimental results on UCI machine learning repository indicated that this method can improve the clustering results with the appropriate attribute weights.