Research on the Influence of Several Factors on Clustering Results using Different Algorithms

Abstract

Clustering analysis is an important technology in the field of pattern extraction and recognition. In order to find the influence of several factors on clustering results using different algorithms and support the decision for power load pattern extraction using clustering techniques, this paper develops the research on the influence of several factors on clustering results using different algorithms. In this paper, three data sets, five normalization methods and five known clustering algorithms including k-means, FCM, SOM, hierarchical clustering and spectral clustering are used, four experiments are designed and performed, they are the influence of the normalization methods on the clustering results, the dependence of clustering results on a data set, the algorithm stability and the sensitivity of clustering algorithm to the input order of the data. The results show that all the factors have obvious influence on the clustering results, and using maximum normalization and FCM algorithm in clustering procedure has the best performance for power load pattern extraction.