基于多种群协同优化的文本分类规则抽取方法  

Rule Extraction Approach to Text Categorization Based on Multi-population Collaborative Optimization

作　　者：刘赫[1,2] 刘大有[1,2] 裴志利[3] 高滢[1,2] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]内蒙古民族大学计算机科学与技术学院,通辽028043

出　　处：《自动化学报》2009年第10期1334-1340,共7页Acta Automatica Sinica

基　　金：国家自然科学基金重大项目(60496321);国家高技术研究发展计划(863计划)(2006AAI0Z245;2006AAl0A309);国家自然科学基金(60773099;60573073);吉林省科技发展计划重大项目(20020303);吉林省科技发展计划项目(20030523);欧盟项目TH/Asia Link/010(111084)资助~~

摘　　要：针对文奉分类中的规则抽取问题,提出一种基于多种群协同优化的文奉分类规则抽取方法.该方法利用信息熵生成初始种群,采用多种群协同优化方法演化当前种群.多种群协同优化方法通过种群之间的相互竞争和良种共享机制提高优化方法的效率.实验结果表明,奉文提出的文本分类规则抽取方法所抽取规则的数量少,准确率高,平均长度短;同时,奉文方法所用的计算时间少,抽取分类规则的速度快,适用于大规模数据集.For the problem of rule extraction in text categorization, a novel rule extraction approach to text categorization based on multi-population collaborative optimization was proposed. Information entropy was applied to generation of initial populations and the multi-population collaborative optimization method was employed to evolve the current population in this proposed approach. The optimization efficiency of this approach was improved by the mutual competition and excellent individuals sharing mechanisms among populations. Experimental results have shown that the number of the rules extracted by this approach is small, and that the accuracy of these rules is high and the average length of them is short. Furthermore, the time of this approach is short and the speed of rule extraction through this approach is high. Therefore, this approach is suitable for large-scale data sets.

关 键 词：规则抽取 文本分类 多种群协同优化 遗传算法 蚁群算法 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]