An Effective Algorithm to Solve Optimal k Value of K-means Algorithm
In spatial clustering, constructing appropriate validity clustering function is crucial to solve optimal cluster number k. The classical K-means algorithm required the k value to be pre-designated. However, it was difficult to designate precise k value in actual applications, which restricted the extensive application of K-means algorithm. An effective algorithm to solve optimal k value, using distance expense function based on Euclidean distance for verifying validity of optimal cluster number, was given in this paper. At the same time, the upper limit of k value was given and the experience rule was reasonably proved. Above all, computed the inside distance and outside distance of clustering, and computed respective distance expense with different k, and then the k value with minimum distance expense was the optimal cluster number. Finally, the algorithm is applied in practice.
Distance Expense Inside Distance K-means Algorithm Outside Distance Spatial Clustering
Taoying Li Yan Chen
School of Economics and Management, Dalian Maritime University Dalian, Liaoning, China
国际会议
杭州
英文
623-626
2006-10-12（万方平台首次上网日期，不代表论文的发表时间）