Modified great deluge algorithm based auto associative neural network for bankruptcy prediction in banks

In the bankruptcy prediction literature, researchers proposed various techniques viz., neural networks, regression analysis, etc. But, in practice, the number of bankrupt banks is very low compared to that of nonbankrupt banks, at least in the Indian context. This induces the problem of data imbalance. To overcome this problem Baek and Cho [1] proposed an Auto Associative Neural Network (AANN) trained with one class i.e. solvent banks data. We proposed and implemented a variant for Baek and Cho's neural network and named it Modified Great Deluge Algorithm based Auto Associative Neural Network (MGDAAANN), wherein a metaheuristic is used to train the auto associative neural network. The efficacy of the MGDAAANN is demonstrated on three bankruptcy data sets taken from the literature. Results indicate that MGDAAANN is a viable single-class classifier to predict bankruptcy in banks with high accuracy. Further, some of the MGDAAANN variants outperformed popular twoclass classifiers such as RBF, orthogonal RBF, MLP, SVM and ANFIS.