Segmentation Region Density Clustering Algorithm Based on Minimum Spanning Tree

To solve the problem that the traditional density clustering algorithms have poor adaptability to imbalanced data sets due to the use of global variables, a density clustering algorithm based on minimum spanning tree is proposed. Firstly, a data set density peaks calculation is used to estimate global density. Secondly, density clustering aims to separate the high-density clusters and low-density area. Thirdly, the minimum spanning tree is constructed and segmented to mining the associations within low density areas, and construct interconnection between high density areas and low density areas. Finally, compute all clusters' density as feature of merging the clusters, and obtain the result. This algorithm combines the knowledge of graph theory, processing the data set by segmentation and combination according to density feature, so that overcomes the limitations of traditional density clustering algorithms. By selecting multiple imbalanced artificial data sets and UCI data sets for test, we verify the effectiveness and robustness of this algorithm. © 2019, Beijing China Science Journal Publishing Co. Ltd. All right reserved.