A Modified Clustering Algorithm Based on Swarm Intelligence

A modified clustering algorithm based on swarm intelligence (MSIC) is proposed in this paper.To improve the running efficiency of the SIC algorithm, the random projection of the patterns into the plane is modified. The patterns are firstly analyzed by principal component analysis (PCA) and the first two principal components (PCs) are retained. The patterns are projected into the plane according to their corresponding PCs, which are processed as the projection coordinates. This modification ensures that the pattern will be similar to the ones in its local surroundings and the rough clustering has been formed at the beginning time of the algorithm. Moreover, to reduce the influence of the parameters on the algorithm, a simple way to calculate the swarm similarity of the pattern is presented. The adjusting formula of the similarity threshold is also proposed. Finally, the modified algorithm is compared with the original one and the results prove the efficiency has been improved significantly.