基于强化型精英保存遗传策略的聚类方法研究  

A Clustering Method Based on the Enhanced Genetic Elite-preservation

作　　者：陈磊[1] 普运伟 

机构地区：[1]昆明理工大学信息工程与自动化学院 [2]昆明理工大学计算中心,云南昆明650500

出　　处：《软件导刊》2017年第12期15-18,共4页Software Guide

基　　金：国家自然科学基金项目(61561028)

摘　　要：遗传算法是一种随机搜索算法,适用于解决许多复杂的智能优化问题。然而,经典遗传算法具有收敛速度慢和易早熟缺陷。为了找到一种普适性高且效果好的改进遗传算法,解决数据聚类问题,提出一种新的遗传算法改进策略。该策略同时保留父代及交叉产生的个体中的绝大部分精英,用来替换掉变异后同等数量的最差个体,并且将交叉与变异概率提高到1,这样不仅能很好地保留住已产生的精英个体,引导算法稳定地向最优解进化,还可最大限度地使算法获得开拓新的解空间能力。实验结果表明,该方法具有较高的聚类准确性和收敛率,平均收敛准确率为94.67%,平均收敛率为100%,且收敛速度较快,是一种适合解决数据聚类问题的可行方案。Genetic algorithm is a stochastic search algorithm,which is applied to solving many complex intelligent optimization problems.However,the classical genetic algorithm has disadvantages of slow convergence speed and easy prematurity.In order to find an improved genetic algorithm with high universality and good effect to solve general data clustering problems,this paper proposes a new revised strategy about GA,which retains parts of the elites in the parents and in the individuals after crossover and increases the probability of cross and mutation to 1.Through these strategies,the novel improved GA makes the algorithm maximize the ability of exploiting new solution spaces,and retains the generated elites well.The experimental results show that,the proposed method has higher clustering accuracy and rate of convergence than two recent improved genetic strategies,whose average accuracy rate achieves 94.67% and average rate of convergence is 100%.In addition,the convergence speed of the proposed method is so fast that it can be as a feasible solution to solve data clustering problems.

关 键 词：改进遗传算法 精英保存策略 数据聚类 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]