A partition clustering algorithm for transaction database based on fcm

Abstract

Large institutions usually have accumulated lots of transaction databases, which refer to multi-database. The effective method for acquiring useful knowledge from the multi-database is to classify them first, and then to mining. Usually, the technology of multi-database classification includes classifying and clustering. This article proposes a partition clustering algorithm to classify multi- database based on FCM. In the algorithm, a membership degree matrix is constructed firstly. And then, in the process of classifying, adjust the matrix to obtain a desired clustering result. Experiments show that our method is reasonable and effective.