Forecasting the Term Structure of Interest Rates Using Neural Networks

Abstract

The term structure of interest rates holds a place of prominence in the financial and economic world. Though there is a vast array of literature on the issue of modeling the yield curve, there is virtually no mention of the issue of forecasting the yield curve. In the current chapter, we apply neural networks for the purpose of forecasting the zero-coupon yield curve. First the yield curve is modeled from the past data using the famous Nelson-Siegel model. Then, forecasting of the various parameters of the Nelson-Siegel yield curve is done using two different techniques: the multilayer perceptron and the feed-forward network. The forecasted Nelson-Siegel parameters are then used to predict the yield and the price of the various bonds. Results show the superiority of the feed-forward network over the multilayer perceptron for the purposes of forecasting the term structure of interest rates.