A Weighted Ensemble Learning by SVM for Longitudinal Data: Turkish Bank Bankruptcy

Abstract

Support Vector Machines (SVMs) are one of the most popular classification methods developed in recent years which have various application fields of research with diverse types of data sets. Longitudinal type of data is one of these types where a great deal of attention should be paid before applying SVMs. In this study, we modeled the decision maker with the idea of ensemble learning on longitudinal financial ratios to discriminate between weak and strong banks validated on Turkish commercial banks data where SVM is considered to be the base learner. We used the success status of the banks as the dependent variable and the financial ratios as independent variables. The results are compared in terms of the modeling performances and sensitivity measures which show the robustness of the model for finding positive instances, i.e. weak banks. The results show that ensemble learning performs better than a single learner. Moreover, we also validated that applying an appropriate normalization technique has strong effects on the performance of the learning step, especially when dealing with longitudinal data.