适用于大数据的遗传优化算法研究

针对传统遗传算法对高维度数据或大数据易陷入局部最优的问题,提出了一种基于网格小生境与分级多种群共同演化的增强遗传算法.首先,采用基于网格的小生境算法建立主种群,主种群独立进化,将低适应度的样本迁移至子种群中.按照低适应度样本的适应度范围建立分级的子种群结构,各子种群内的样本独立演化,低适应度子种群的样本可进化并迁移至高适应度种群或返回主种群,从而防止具有一定竞争力的样本过早死亡.对比实验结果表明,本算法对高维度数据具有较好的优化效果,优于同类型遗传算法.

Evolutionary Optimization Approach Research for Big Data

For the problem that conventional evolutionary approach is easy to trap local optimal for high di-mensionality or big data ,an enhanced evolutionary algorithms based on grid niches and multi-layer popula-tion has been proposed .Firstly ,based on grid niches approach the main population is constructed ,main population evolution independently and migration the members with low fitness value to sub-populations . By the fitness range of the low fitness individuals the sub-population constructed ,each sub-population evo-lution independently ,low fitness individuals in the sub-population could migrate to main population ,with that operations the diversity is produced and the premature convergence is prevented for big data .Compared evalu-ation for benchmark problems result show s that the proposed approach has superior performance .