摘要
文本特征选择对提高文本分类的速度和准确率,改善网络信息过滤效果至关重要.把特征选择看作优化组合问题,提出用遗传算法进行文本特征选择.传统遗传算法适应性较差,本文对传统遗传算法交叉概率、变异概率、更新策略等重要参数和关键环节作了改进,实验验证了该算法的有效性.
Text feature selection is crucial to improve the speed and accuracy of text categorization, and then improves the effect of network information filtering. Considering feature selection as optimizing problem and in view of traditional Genetic Algorithm＇ s shortcomings, we improve important parameters and core steps of general genetic algorithm, such as self - adapting ability of crossover probability, mutation probability and selection operator. Simulating result illustrates validity of the improved genetic algorithm.
出处
《山东师范大学学报（自然科学版）》
CAS
2007年第2期17-19,共3页
Journal of Shandong Normal University(Natural Science)
关键词
信息过滤
特征选择
遗传算法
information filtering
genetic algorithm
feature selection
