Bankruptcy Prediction in Banks by an Ensemble Classifier

Abstract

In this paper we developed a set of ensemble classifiers using simple majority voting scheme. As part of the ensemble, we used seven classifiers viz., ANFIS, SVM, Linear RBF, Semi-online RBF1 and Semi-online RBF2, Orthogonal RBF, MLP. We designed the ensembles by taking two, three, four, five and six classifiers at a time from the seven classifiers. In each case we selected the combination that gave the highest classification rate and least Type-I error. We used the two well-known bankruptcy data sets viz., (i) Spanish banks data and (ii) US banks data for the study. The models ANFIS, Semi-Online RBF2 and MLP emerged as the most important models as they figured in the best ensemble combinations.