摘要：As information technology is developing rapidly, massive and high dimensional data sets have appeared in abundance. The existing attribute reduction methods are encountering bottleneck problem of timeliness and spatiality. AP(Affinity Propagation) is an efficient and fast clustering algorithm for large dataset compared with the existing clustering algorithms. This paper discusses attribute clustering method in order to reduce attributes and provides a kind of parallel attribute reduction algorithm based on Affinity Propagation （APPAR） clustering. The attribute set is clustered into several subsets by Affinity Propagation algorithm first, and then the reductions of these subsets are proposed concurrently in order to get attribute reduction set of the whole data set. The whole algorithm has been improved in the two sides so as to largely increase the algorithm’s speed. Experimental results show that the APPAR method is outperforming traditional attribute reduction algorithm for huge and high dimensional dataset processing.
