Towards Interactive Clustering on Parallel Environment

Clustering is one of the major data mining applications. An obvious characteristic of data mining distinguished from traditional data processing is that the conclusion of data mining cannot be predicted. Data mining is a multi-step process, and user must be allowed to be the front and the center in this process, especially clustering mining method. In this paper, the necessity of interactive data mining is illustrated. A framework of high performance interactive data mining on PC-cluster is proposed, and an interactive clustering algorithm for multidimensional data on the framework is presented.