论文部分内容阅读
针对大量无关或冗余的特征通常会降低模式分类中分类器性能的问题,提出一种基于异步并行微粒群优化的特征子集选择方法(AP-PSO).该方法采用二进制微粒群优化搜索特征子集,利用异步并行方式提高算法的运算效率;为有效协调种群的全局探索和局部开发能力,充分利用混沌运动的遍历性和随机性,提出一种一致混沌变异算子.与已知4种特征子集选择方法进行比较,所得结果验证了该算法的有效性.
Aiming at the problem that a large number of irrelevant or redundant features usually reduce the performance of classifiers in pattern classification, a feature subset selection method (AP-PSO) based on asynchronous parallel particle swarm optimization is proposed. This method uses binary particle swarm optimization In order to effectively coordinate the global exploration and local development ability of the population and make full use of ergodicity and randomness of chaotic motions, a uniform chaotic mutation operator is proposed, which is similar to the known four kinds of chaotic mutation operators Feature subset selection method to compare the results obtained verify the effectiveness of the algorithm.