Confusion relationship between patterns and its application in adaptive construction of hierarchical classifiers

The hierarchical relationship and the objective patterns of subclassifiers is the primary difficulty to construct a hierarchical classifier. In order to solve this problem, firstly, the confusion relationship between patterns has been defined to describe the interweaving effects of patterns in the decision domain. Then a measurement of the relationship has been proposed by utilizing the confusion matrix. Abiding by the Fisher Principle, a multipattern confusion relationship analysis machine (MPCRAM) has been designed to adaptively construct the structure of a hierarchical classifier. Various data scenarios have been used to compare the hierarchical structures generated with the MPCRAM and the conventional ways. The results have testified that MPCRAM was effective, and it could prominently improve the performance of a hierarchical classifier.