The Extension Of Proportional Association Measure

The local-to-global proportional association measure proposed by Huang-Shi-Wang (2013) is based on without considering revenue weight in response variables and cost in the explanatory variables. But in the actual application, question with cost and revenue information often, and we need to make a optimal decision according to an association measure with cost and revenue. Basing on the local-to-global proportional association measure proposed by Huang-Shi-Wang, firstly, an association measure is proposed to maximize the expected ROI. Secondly, many practical application with misclassification costs information, an association mea-sure is also proposed to maximize the expected profit with misclassification matrix.Meanwhile, a descriptive experiments with a synthetical data set is presented. At last, a modified proportional prediction model was proposed, and it enhance pre-diction accuracy rate under a low distribution bias.