Classification with Choquet Integral with Respect to Signed Non-additive Measure

In order to better understand the nature of classification, a data modeling-based perspective is needed. When the attributes in the database have high interactions that make the non-linear relationships, the use of linear model as the aggregation tool for data modeling is not appropriate. With this consideration, in this paper, we studied the Choquet integral with respect to signed non-additive measure to aggregate the data and proposed a new classification method. We discussed the basic idea and mathematical framework of the non-additive measure and its geometric meaning. Based on the theoretical works, we conducted an experimental test by comparing our approach with others on a real life classification problem on credit card holders´ data set with high dimensionality was shown to demonstrate the effectiveness and efficiency of the proposed approach.