基于类频繁模式树的关联分类

Algorithm for Mining Associative Classification Rules Based on Class Frequent Pattern Tree

提出一种新的基于类频繁模式树的关联分类算法CFPC(Class FP-tree based Classifier).该方法基于FP-tree实现,无需生成庞大的候选项目集;依据记录的分类属性进行指导性划分,并使用类支持度进行记录项的分类剪枝,生成类模式树,避免了小数据类别集上的强关联模式遗漏;挖掘出的规则形成分类器,用于类标号未知的记录的区分.试验结果表明CFPC的正确性和有效性.

This paper presents a class FP-tree based associative classification method, named CFPC. We adopt the method based on FP-growth, partition data sets according to class attribute of each record, and construct class FP-tree . It applies an efficient pruning method based on minimal support of each class to keep correct and high quality rules to form classifier. Such a process can avoid loss of strong frequent pattern in small class sets; unlabeled records are classified using rules founded by association rule mining algorithm. Experiments indicate that CFPC is correct and efficient.