基于概化闭包压缩规则的关联分类器构建算法研究

Research on Algorithm Building of Associative Classifier Based on Generalized Closed Compact Rules

在分析关联规则挖掘领域中概化闭包(GC)项集压缩方法基础上,为克服训练数据集中的噪声干扰,改进L3G分类器,设计了一种基于概化闭包压缩规则的关联分类器(ACGCCR)构建算法模型。ACGCCR改进了GC的容忍限度值设置方法,根据类分布状态自适应设置容忍限度值;并规定一种新的概化闭包类规则裁剪方法,避免概化过程中出现学习能力不足的问题。ACGCCR分类规则在压缩存储表现、预测准确度、算法鲁棒性等方面性能表现良好。

In order to overcome noise interference in training data set, an improved LG^3 classification and an algorithmic model of Associative Classification Based on Generalized Closed Compact Rules（ACGCCR）are proposed. ACGCCR improves tolerance factor setting method of generalized closed rules mining, providing an adaptive tolerance factor setting method according to class distribution state. A new generalized closed class rules pruning technique is put forward to avoid insufficient learning ability problem. ACGCCR classification rules have better compact storage performance, stronger predictive accuracy and algorithmic robust capability.