On resources optimization in fuzzy clustering of data streams

Abstract

In this paper the resource consumption of the fuzzy clustering algorithms for data streams is studied. As the examples, the wFCM and the wPCM algorithms are examined. It is shown that partitioning a data stream into chunks reduces the processing time of considered algorithms significantly. The partitioning procedure is accompanied with the reduction of results accuracy, however the change is acceptable. The problems arised due to the high speed data streams are presented as well. The uncontrolable growth of subsequent data chunk sizes, which leads to the overflow of the available memory, is demonstrated for both the wFCM and wPCM algorithms. The maximum chunk size limit modification, as a solution to this problem, is introduced. This modification ensures that the available memory is never exceeded, what is shown in the simulations. The considered modification decreases the quality of clustering results only slightly. © 2012 Springer-Verlag Berlin Heidelberg.