Classification Based on Average Length of JEPs
基于JEP平均长度的分类方法

JEP is a kind of itemset which support in one dataset is zero but in another dataset not zero. After studying the problems of JEPs in the classification procedure, we presents the concept of independent supports of itemsets. Compared with traditional supports, independent supports provide more detailed distribution information of the dataset and more powerful classifiers can be built on them. Then we propose the concept of average length based on independent supports of JEPs. Finally, we present a classification algorithm using the average length of JEPs as the classification feature. This algorithm can give a more precise classification to data on the border of the datasets.