Performance Evaluation of ECOC Considering Estimated Probability of Binary Classifiers

Error-Correcting Output Coding (ECOC) is a method for constructing a multi-valued classifier using a combination of the given binary classifiers. ECOC is said to be able to estimate the correct category by other binary classifiers even if the output of some binary classifiers is incorrect based on the framework of the coding theory. Although it is experimentally known that this method performs well on real data, a theoretical analysis of the classification accuracy for ECOC has yet to be conducted. In this study, we evaluate the superiority of a code word table in showing the combinations of binary classifiers of ECOC that have been experimentally demonstrated. In other words, we analytically evaluate how the estimation of the categories is influenced by the estimated posterior probability, which is the output of the binary classifier, as well as by the structure of constructing the code word table.