基于改进混合蛙跳算法的粗糙属性交叉熵优化约简

Rough attribute reduction with cross-entropy based on improved shuffled frog-leaping algorithm

结合粗糙集属性约简二进制优化模型,提出一种基于改进混合蛙跳算法的粗糙属性交叉熵优化约简算法,该算法将粗糙集属性划分至不同蛙群进化模因组内,每个模因组内属性集设计成以精英个体为中心力的蛙群并行演化方式,并采用交叉熵最小原理进行精英个体寻优全局最优约简集,快速而有效地处理大规模信息系统的属性约简.UCI仿真实验结果表明本文提出的算法在搜索全局最小属性约简解效率和精度方面具有明显优势,该算法应用于含噪音的人脑核磁共振图像MRI分割实验,其对MRI图像分割的高效性进一步表明该算法具有较强的适用性.

According to the binary optimization model of attribute reduction in Rough Sets,a novel and efficient rough attribute reduction with cross-entropy based on improved shuffled frog-leaping algorithm （named SFCEAR）was proposed in this paper.In the proposed algorithm,the rough attribute set was grouped into different frogs evolutionary memeplexes.An especial kind of frog structure with the central force as their respective evolutionary elitist was designed in each attribute memeplex space,and the minimal principle of cross-entropy method was adopted to carry on the elitist frog optimization for attribute reduction.The algorithm could make efficient reduction for an information system with a large number of attributes.Experimental results on UCI data sets demonstrated the proposed algorithm outperformed other existing similar algorithms in both efficiency and accuracy of minimum attributes reduction.Moreover the proposed approach is applied into magnetic resonance images（MRI）segmentation, and the effective and robust segmentation results further indicate it has stronger applicability.