146 Clustering Algorithm Study Based on HTS-SOM

The paper presents a new Self-Organizing Map (SOM) called HTS-SOM (HashTree Structure-SOM). This network can grow automatically, visualize data efficiently, query quickly, and possess a clear hierarchy structure. It overcome the restrict of which the tradition SOM model must be appointed in advance and can not hierarchically cluster, especially solve the problem that the time-consuming search for the Best Matching Unit in large maps. Experiment with real world data showed that the HTS-SOM is noticeably better in performance than the SOM and TS-SOM. It has computational complexity relative to the TS-SOM, but the topology is flexible and the search effectively.