一种求解类覆盖问题的混合算法

A Hybrid Algorithm on Class Cover Problems

提出一种扩展的类覆盖问题,并将它归纳为一个有约束的多目标优化问题模型,该问题的解决对构建强壮的分类识别系统具有重要的意义.因此,通过对二进制粒子群算法参数特性的深入分析,阐明二进制粒子群算法不仅具有良好的全局搜索特性,而且能够充分利用已有的先验知识.进而提出一种贪心算法与二进制粒子群优化算法相结合的混合算法求解扩展的类覆盖问题,该算法在获得更优解的同时,仍具有较快的运算速度.多种算法的比较结果表明了算法的有效性和可行性.

An extended class cover problem is presented and then it is reduced to a constrained multi-objective optimization problem. Solving this problem is significantly important to construct a robust classification system. Therefore, through analyzing the parameters of the binary particle swarm optimization, the conclusion that the binary particle swarm optimization can not only explore the search space efficiently, but also utilize the apriori knowledge adequately, is drawn in this paper. Furthermore, a hybrid algorithm combined with the conventional greedy algorithm and binary particle swarm optimization algorithm is proposed to deal with the extended class cover problem. The proposed algorithm can get a better solution in less runtime and the simulated comparative results with other algorithms show its feasibility and validity.