摘要
针对目前客户流失分析中的过量抽样和规则不容易理解导致分类算法在实际应用中的效果不明显的问题,提出了一种抗体抗原交叉的规则归纳算法(IRAA)。该算法基于人工免疫思想,结合了Michigan方法模型的规则提取和分类方法,并且在其抗体的克隆选择过程中增加抗体与抗原的交叉。通过与传统分类算法的比较实验表明,IRAA提高了分类准确率,导出了容易理解的规则。
Concerning to the low efficiency of the classification algorithm in the real applications which was caused by over-sampling and misunderstanding rules in the analysis of customer loss, a new classified method, Induction of Rule with Antibody-cross-antigen of Artificial immune system （IRAA）, was proposed. IRAA was based on artificial immune system, and combined with both rule extraction and classification of Michigan approach model. Furthermore, antibody-cross-antigen was added in the antibody＇ s clone selection process. Compared with tradition classification algorithms in actual application, results show the proposed algorithm obtains higher precision and easy-to-understand rules.
出处
《计算机应用》
CSCD
北大核心
2008年第7期1705-1708,共4页
journal of Computer Applications
关键词
客户流失分析
数据挖掘
分类
人工免疫系统
克隆选择
customer loss analysis
data mining
classification
artificial immune system
clone selection
