A Novel Approach to Identifying Global Exceptional Patterns in Distributed Data Mining

Abstract

With the increasing development and application of distributed database, distributed data mining has attracted many data mining researchers’ attention. In this paper, a framework for distributed data mining is introduced, and based on the framework, many patterns are generated from each database after data mining, so it is necessary to synthesize all the patterns to identify the meaningful global patterns. An approach to synthesizing local patterns to identifying global exceptional patterns is developed. In this approach, a pattern’s significance is measured by the deviation of the pattern’s support from the average support. Experimental results show that our approach is reasonable and appropriate to identify exceptional patterns.