Parametric and nonparametric evolutionary computing with a content-based feature selection approach for parallel categorization

摘要

This paper proposes two modified evolutionary computing methods for genetic algorithms (GAs) and proves an effective content-based feature selection approach to improve clustering performance. The conventional GAs suffer from the problem of slow learning and are prone to be trapped into a local minimum due to a high dimensional exploration space. In this paper, we propose a parametric and a nonparametric evolutionary algorithms to properly adjust the operators of GA. In the parametric approach, several fuzzy control parameters are artificially defined to adaptively optimize the GA behaviors. By contrast, they are automatically adjusted by GA itself in the nonparametric approach. Moreover, a content-based feature selection (CFS) approach is demonstrated to create a robust semantic space and reduce the number of dimension which accelerates the speed of evolutionary computing. We take advantage of a parallel computing technology to improve the efficiency of clustering. The experimental results show that our methods enhance the performance of the standard GA and are more efficient than those implemented on a single processor. The CFS approach not only reduces the document dimension, but also indirectly advances clustering efficiency.