融合变精度粗糙熵和协同进化的概念格挖掘算法

Concept lattice mining algorithm using rough entropy with variable precision thresholding and co-evolution

为解决概念格挖掘优化问题,借鉴变精度粗糙集模型和协同进化思想,提出了融合变精度粗糙熵和全局粒子群的概念格协同挖掘算法(REVPT)。该算法引入变精度粗糙熵对各概念格子群动态度量建立粗糙近似格,并通过种群之间协作共享寻优经验提高概念格的全局挖掘优化能力,有效缩减原格群规模并挖掘出一致粗糙分类规则。实验结果表明,当变精度粗糙熵阈值β处于某一合适范围,该算法在保证收敛速度同时具有较强的全局建格优化能力,在知识挖掘精度和效率方面具有较好的鲁棒性。

Based on some special advantages of variable precision rough sets model and co-evolutionary particle swarm algorithm,a novel concept lattice mining algorithm（REVPT）using rough entropy with variable precision thresholding and co-evolution was proposed to solve some optimization problems of the concept lattice mining.In this algorithm,variable precision rough entropy was used to scale the subpopulations of various concept lattices dynamically,and rough approximation lattices constructed.The global optimization efficiency of concept lattices was improved by sharing search experiences among different populations,which can reduce the scale of the former concept lattices,and deduce the consistent decision rule sets efficiently.The experimental results show that the proposed algorithm is better on the convergence and lattice optimization when the variable precision thresholding β is at a certain appropriate range.Therefore it is of robustness on the rules mining accuracy and efficiency.