Classification of Domestic and Foreign Commercial Banks in Turkey Based on Financial Performances Using Linear Discriminant Analysis, Logistic Regression and Artificial Neural Network Models

Abstract

The Data Mining (DM) techniques of linear discriminant analysis (LDA), logistic regression (LR) and artificial neural network (ANN) models are among the multivariate techniques used for predicting the predefined class membership of dependent variables. Hence, the aim of this study is to discuss and illustrate LDA, stepwise LDA, LR, forward LR and four types ANNs and compare these models’ correct classification ability. For this purpose, the data of commercial banks operating in Turkey in two pre-defined groups, namely domestic and for- eign banks, is used. In this study, the classification performance of ANN models against to LDA and LR is investigated. The ability of these classification methods in classifying the banks correctly is compared in terms of correct classification rates. As the results reveal that ANN (ANN-Prune) outperforms LDA and LR in terms of bank classification accuracy and thus, provide an effective alternative for implementing bank classification.