A novel attribute reduction algorithm based on peer-to-peer technique and rough set theory

Abstract

Rough Set theory is an effective tool to deal with vagueness and uncertainty information to select the most relevant attributes for a decision system. However, to find the minimum attributes is a NP-hard problem. In this paper, we describe a method to decrease the scale of the problem by filtering core attributes, and then employ the checking tree to test the rest attributes from bottom to top by using peer-to-peer technique. Furthermore, we utilize pruning method to enhance the speed and discard the node when one of its child node superset of certain attribute reduction found before. Experimental results show that our parallel algorithm has the high speed-up ratio while the attribute reductions are distributed in the bottom of the tree. In a peer-to-peer network, our algorithm will amortize the required memory on client computers. Accordingly, this algorithm can be applied to deal with larger data set in a distributed environment.