A Framework of Mixed Model of a Neural Network and Boolean Logic for Financial Crisis Prediction

Abstract

AbstractThis paper proposes a framework for pattern classifications. This framework includes two neural networks and a logic circuit. The first neural network serves to classify regular instances, and the second one handles the overlapped instances caused by the first neural network. The proposed framework integrates the two neural networks by the logic circuit and then makes the final classifications. We provide two methodologies using the proposed framework including financial crisis prediction and principle components analysis. In the financial crisis prediction case, we use the financial ratios of individual companies as the inputs of the first neural network and the macroeconomic indicators as the inputs of the second neural network. In the principle components analysis case, we use the attributes in a major factor group as the inputs of the first neural network and those in a minor factor group as the inputs of the second neural network.