Comparative study of clustering methods based on linear data distribution

Based on different linear data distribution patterns in a two-dimensional space, constructing two kinds of artificial simulated linear data distribution patterns in a three-dimension space. Three clustering methods are presented and discussed by comparative experimental analysis model. The results by using three kinds of different clustering methods which are K-means method, Two-step method and Kohonen method visually illustrate the different clustering results. In the end of this paper, we could come to the conclusion that the applicability of different clustering methods on three-dimensional spatial linear data distribution. Clearly, when the kind of three-dimensional spatial linear data distribution is simple, at the same cluster number case, the clustering evaluation score of single method is higher. That is to say, the single method is more suitable to simple linear data distribution, especially Two-step clustering method is the priority selection. By comparison, when the kind of three-dimensional spatial linear data distribution is hybrid, the combination method is more suitable. The more optimized clustering analysis process is presented based on the above comparative analysis.