基于互信息的改进量子蛙跳算法属性约简

Attribute Reduction of Improved Quantum-SFLA Based on Mutual Information

对零售行业的信用评价属性约简进行了研究。利用考虑删去某个条件属性后条件属性与决策属性的互信息是否改变来计算条件属性的属性核。为保障种群的多样性,利用混沌算法对量子蛙的叠加态的观测态赋值及二进制编码,对量子蛙进行适应度评价。按适应度大小对青蛙进行种群分配,对每个种群中最劣青蛙对最优青蛙进行改进步长的量子旋转门学习。经过一定迭代次数后,将所有种群的青蛙进行混合之后学习,再按适应度对青蛙进行种群分配。经过一定种群内学习及种群间学习后使其最优青蛙与最劣青蛙的适应度值之差小于阈值停止迭代从而达到属性约简的目的,从而得出零售行业信用评价的最简属性。

This paper studies on the attribute reduction of credit rating in the retail industry.Firstly,by deleting a condition attribute and then judging whether the mutual information between condition attributes and decision attribute change or not to calculate the core condition attributes.To guarantee the diversity of the population,we use chaos algorithm to both encode the observation state of superposition state of quantum frogs and make the binary encoding.Quantum frogs＇ fitness evaluation was performed,and frog populations were assigned to each population according to the fitness evaluation value.The worst frog learned improved step quantum revolving door from the best one in the population.After a certain number of iterations,the frog populations were all blended learning,and then allocating the frog populations according to fitness value of each frog.After a certain number of times of internal population study and among populations study,the difference of the fitness value between the best frog and the worst one was less than the threshold value so as to achieve the purpose of stopping the iteration and got the attribute reduction,thus obtaining the most simple attributes of the retail industry credit evaluation.