(The) prediction of interest rate using artificial neural network models

Abstract

Most previous researches for the prediction models of interest rate have made efforts to test empirically whether interest rate determination theories may be applied to Korean situation rather than to forecast interest rate. For the application of forecasting interest rate, those interest rate determination models are not appropriate. Artificial neural network models were used for forecasting interest rate as a new methodology, which has proven itself successful in financial domain. This research intended to construct artificial neural network models which can maximize the performance of prediction, regarding Corporate Bond Yield(CBY) as interest rate. We combined the fundamental variables derived from interest rate determination theory and the market variables considering the supply and demand of corporate bond for the construction of models. While the models which consist of only time series data for corporate bond yield were developed, the other models generated through conjunction and reorganization of fundamantal variables and market variables were developed. Every models were reconstructed to predict 1, 3, 6, 12 months after and we obtained 16 artificial neural network models for interest rate forecasting. The 132 neural networks were learned for searching optimal model. Multi-layer perceptron networks using backpropagation algorithm showed good performance in the prediction for 1 month after. The RMSE was through 0.238 to 0.593 and artificial neural network models were better performance with 5% significant level by t-test. In the prediction for 3 months after, the combining model with the fundamental and intermarket varibles(NN4) showed significantly the good performance in forecasting the interest rate. The determination of how long period to forecast is very important factor in constructing the interest rate forecast models.