一种基于正交法和扩展卡方检验的关联分类算法

针对几种典型分类算法中存在的诸如分类器性能较低和算法效率不高等问题，提出了一种基于正交法和扩展χ^2检验的分类算法ERAC。算法首先通过正交法产生所有的频繁项集和关联规则，然后采用一种扩展χ^2检验来对规则进行分级和修剪，有效减少分类器的规则数目。试验结果表明，该算法与CBA等算法相比较具有较高的分类准确率和运行效率。

Multi-class associative classification based on intersection method and extended chi-square testing

Given that there exit some defects in several typical classification algorithms, such as poor classification performance and long running time when the algorithm efficiency is not high, this paper proposed a classification algorithm ERAC based on intersection method and extended chi-square testing. This algorithm first produced, through intersection method, all the frequency items and association rules. Then it conducted classification and pruning on rules using an extended testing method, reducing the number of rules for classification effectively. Subsequent experiments prove that new method, compared with the CBA algorithm, has higher classification accuracy and operating efficiency.