基于混沌集群智能优化算法的多目标粗糙集属性约减

Attribute Reduction of Multi Objective Rough Set Based on Chaotic Swarm Intelligent Optimization Algorithm

针对管理实践及大数据处理过程中具有多决策属性的粗糙集属性约减问题,将条件属性依赖度与知识分辨度进行结合构建属性权重,分别建立针对不同决策属性的约减目标函数,引入帕累托最优思想,将基于多决策属性的粗糙集属性约减问题转化为离散多目标优化问题。针对该问题的结构设计了具有集群智能优化思想的元胞自动机求解算法,在算法中引入基于个体的非支配解集平衡局部最优与全局最优的关系,引入混沌遗传算子增加种群多样性。以某铁路局设备安全风险处理数据为案例构建多决策属性粗糙集决策表进行优化计算并进行管理决策分析。研究发现:(1)相对于传统的NSGA-II与MO-cell算法,本文提出的算法具有更强的多目标属性挖掘性能;(2)帕累托最优思想可以较好地解释多决策属性粗糙集在管理实践中的意义。

For the problem of attribute reduction of rough set with multi decision attributes in the management and big data processing,objective functions for different decision attributes are established based on the dependability and knowledge resolution of condition attributes.By introducing the idea of Pareto optimal solution,the attribute reduction problem of rough set with multi decision attributes is converted into multi-objective optimization problem with discrete variables.Based on the idea of swarm intelligence,the new cellular automata solution algorithmis designed,and in the algorithm,the set of individual non-dominated solutions is designed to balance the local optimum andtheglobal optimum.Moreover,chaoti coperator is introduced to increase the population diversity.Inthe numerical example,the decision tables are established,optimized and analyzed based on the security risk handling data of arailway company.The study finds that(1)compared with the traditional NSGA-II and MO-cell algorithms,the new algorithm shows better performance in multi-objective attribute mining;(2)the idea of Pareto optimal solution can well explain the application of rough set with multi decision attributes in management.