改进的遗传算法在分类规则挖掘中的应用

The Application of Improved Genetic Algorithm in Classification Rule Mining

与其它常用的分类方法相比,遗传算法具有较强的伸缩性和全局搜索能力。本文对分类规则进行二进制编码,并通过适应度函数来评价分类规则的有效性。文中对简单遗传算法进行了改进,并引入相似度的概念,提出基于相似度的交叉算子。首先设定一个相似度阈值,计算个体的相似度与相似度阈值比较,若大于该阈值,则执行均匀交叉操作,否则执行单点交叉操作。最后采用USI机器学习数据库中的数据进行实验。实验结果表明改进的遗传算法挖掘出的分类规则准确率较高。

Compared to other commonly used classification , the genetic algorithm is superior scalability and global search capabilities. In this paper,effectiveness of classification rules for binary encoding is evaluated with the fitness function.Simple genetic algorithm has been improved,and the concept of similarty is introduced,and the crossover operator based on the similarty is proposed in the paper.First,we set a similarty threshold value and calculate the individual ’s similarity,which is compared with the similarity threshold value,if it is greater than the threshold,then the uniform crossover operator is implemented,or the single-point crossover is implemented.At last we do the experiment with the data in USI machine learning database.The results indicate that improved genetic algorithm excavated a high accuracy rate of classification rules.