A Hybrid Model On Data Clustering And Computational Intelligence For Bank Crisis Classification And Prediction

Abstract

The prediction of bank failures is an important academic topic of which many have used artificial intelligence methods to build an early warning system for this purpose. The objective of this study is to enhance the accuracy in predicting bank failures by proposing two hybrid models. In this research, two hybrid models are developed by integrating a K-means cluster method to integrate K-means with a Back-Propagation Neural Network (BPN) and a Support Vector Machines (SVM) technique for financial data classification. Datasets from the website of Federal Reserve Bank of Chicago are employed for benchmark test. Initially a K-mean clustering method is applied to preprocess the dataset thus a more homogeneous data within each cluster will be attainted. A clustering method is employed to separate the case library into smaller clusters; and lastly, a Back-Propagation Neural Network (BPN) and a Support Vector Machine (SVM) model are established and prediction results are being generated. The average forecasting accuracy for bank failures of Kmeans-BPN model is 92.79% and Kmeans-SVM model is 92.43%. In comparison to other methods, the proposed model outperforms other prediction models as the prediction accuracy of bank failures are being enhanced while it simultaneously produces valuable information for business owners and investors.