基于混沌离散粒子群的粗糙集属性约简算法  

Rough Set Attribute Reduction Algorithm Based on Chaotic Discrete Particle Swarm Optimization

作　　者：栾雨雨 王锡淮[1] 肖健梅[1] LUAN Yu-yu;WANG Xi-huai;XIAO Jian-mei(College of Logistics Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学物流工程学院,上海201306

出　　处：《计算机仿真》2021年第7期271-275,共5页Computer Simulation

基　　金：国家自然科学基金(61573240)。

摘　　要：针对粒子群属性约简算法容易早熟、易陷入局部最优解的问题,提出一种融合混沌离散粒子群与粗糙集的属性约简算法(CBPSORS)。在该算法中,首先利用混沌序列初始化粒子的位置和速度,得到一个无序的粒子种群。其次改进最优粒子进行混沌变异过程,改进惯性因子和加速因子来提高算法性能。再次用粗糙集理论对生成的属性子集相关性进行评估。最后用K-近邻(KNN)算法生成分类模型在UCI数据集上对该算法进行验证。理论分析与实验结果表明,与基于粗糙集的属性约简算法(RS)、基于粒子群的粗糙集属性约简算法(PSORS)以及基于遗传算法的粗糙集属性约简算法(GARS)相比,文中算法可以在保持决策表知识信息的前提下,约减掉更多的条件属性,提高分类精度。In order to solve the problem that particle swarm optimization(PSO) is easy to premature and fall into local optimal solution, an attribute reduction algorithm(CBPSORS) based on chaotic discrete particle swarm optimization and rough set is proposed. Firstly, the chaotic sequences were used to initialize initial position and the location of particles in order to obtain a chaotic particle population. Secondly, the optimal particle was improved to carry out the chaotic mutation process, and the inertia factor and acceleration factor were improved to improve the performance of the algorithm. Thirdly, rough set theory was used to assess the relevance of the potential generated attribute subsets. Finally, the K-nearest neighbor(KNN) algorithm was used to generate the classification model to verify the algorithm on the UCI datasets. The theoretical analysis and simulation results show that compared with rough set based attribute reduction algorithm(RS), particle swarm based rough set attribute reduction algorithm(PSORS) and genetic algorithm based rough set attribute reduction algorithm(GARS), the CBPSORS algorithm can reduce more conditional attributes and improve the classification accuracy on the promise of maintaining the knowledge information of decision table.

关 键 词：粒子群优化 粗糙集理论 混沌优化 属性约简 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]