- 期刊
- OpenAccess
A Two-stage Clustering Method Combining Ant Colony SOM and K-means
並列摘要
Self-organizing map (SOM), which is an orderly mapping technique, can convert complex, non-linear and high-dimensional data into simple, geometric and low-dimensional data that can easily be visualized. In data analysis techniques, the capability of SOM and K-means for clustering large-scale databases has already been confirmed. Although SOM and K-means have their superior features for cluster analysis, their combination into a two-stage method is generally much more powerful than the two methods used individually. In this research, an ant-based self-organizing map (ABSOM) is proposed. The ABSOM embeds the exploitation and exploration rules of state transition into the conventional SOM algorithm to avoid falling into local minima. To examine the usefulness of the proposed method, the ABSOM is combined with K-means into a two-stage clustering method, i.e. ABSOM+K-means. Applied four public data sets, the ABSOM has been proved that it performs better than Kohonen's SOM and it also works very well in the two-stage cluster analysis when it is taken as a preprocessing technique.