A New Hybrid Approach For Forecasting Interest Rates

The dynamic, non-linear, volatile and complex nature of interest rates makes it hard to predict their future movements. In order to deal with these complexities, the authors propose a two-stage neuro-hybrid forecasting model. In the initial data preprocessing stage, multiple regression analysis is implemented to determine the variables that have the strongest prediction ability. The selected variables are then provided as inputs to a Fuzzy Inference Neural Network to forecast future interest rate values. The proposed hybrid model is implemented using data from the U. S. interest rate market.