关键词: 关联分类, 多次学习, 关联度, 分类准确率

Abstract:

The Association classification algorithms based on support?confidence framework are difficult to generate a large number of high?quality rules; Meanwhile,in several types of data sets like imbalanced data sets,the generated rules cannot cover all training instances,which causes low classification accuracy. To solve these problems,we propose an Improved Algorithm based on Multiple learning and Correlation degree (IAMC). Firstly,when extracting rules,IAMC performs multiple times of associative classification learning on the training set,in order to generate more rules. Secondly,to improve the quality of the generated rules,a new measure of correlation degree is adopted by IAMC,which takes into account the confidence and the CCS degree. Finally,after extracting the association classification rules,IAMC uses the decision tree method to extract the rules from the training instances which cannot be covered by the associative rules. The experimental results show that IAMC algorithm provides more high?quality rules and can also achieve higher classification accuracy than existing associative algorithms on multiple UCI datasets.

Key words: associative classification, multiple learning, correlation degree, classification accuracy