一种结合关联与FOIL算法的分类方法

针对如何减少关联分类方法中冗余规则，增加FOIL算法的规则数，以提高分类准确率，提出了一种结合关联与FOIL算法的分类方法，并称之为ACFA．首先，以类支持度和自信度为度量提取长度为1和2的规则，其次，利用Apriori算法挖掘出频繁2-项集F2，然后在频繁2-项集F2申挑选满足条件的频繁项建立候选集，最后在候选集上运用FOIL算法来产生分类规则．实验表明算法ACFA不但有效减少了关联分类方法中冗余的规则，并大大增加了FOIL算法的规则数，提高了分类的准确率．

A Classification Approach of Integrating Associative Classification and FOIL Algorithm

In order to get high classification accuracy, we need generate less redundant rules in aLssociation classification and increase the .number of rules in FOIL algorithm. Thus, a new method called ACFA （a classification approach of integrating associative classification and FOIL algorithm） is proposed. First, ACFA generates length-1 and length-2 classification rules. Secondl ACFA uses the Apriori algorithm to mine the frequent pairs of items. Third, it selects the frequent pairs which satisfy the given conditions to construct the candidate set. Finally, ACFA adopts the FOIL algorithm to generate classification rules based on the candidate set. Our experimental results show that ACFA generates less redundant rules than association classification. Meanwhile, ACFA generates larger rules than FOIL algorithm. Therefore, ACFA gets high accuracy.