Interclass analysis in symbolic pattern classification problems

Summary

This paper presents a method of interclass analysis in pattern classification problems for symbolic objects. In our problems, each object is described not only by usual numerical features, but also by other type features including interval valued features, nominal features, and ordinal qualitative features. We introduce the Cartesian System Model (CSM) as a mathematical model to treat symbolic objects. Then, we define the notions of the inside view and the outside view based on the neighborhood relations. For a given feature subset, the size of outside view and the size of inside view indicate the interclass distinction and the generality of class descriptions, respectively. Our interclass analysis is realized by combining a simple local feature selection method with the sizes of inside and outside views. Since our interclass analysis is classifier independent, it may be useful as a pre-processing process in the design of many pattern classifiers. Several experimental results are also presented.