Recurrent Artificial Neural Networks (RANN) for forecasting of forward interest rates

Abstract

There are numerous methods for estimating forward interest rates as well as many studies testing the accuracy of these methods. The approach proposed in this study is similar to the one in previous works in two respects: firstly, a Monte Carlo simulation is used instead of empirical data to circumvent empirical difficulties: and secondly, in this study, accuracy is measured by estimating the forward rates rather than by exploring bond prices. This is more consistent with user objectives. The method presented here departs from the others in that it uses a Recurrent Artificial Neural Network (RANN) as an alternative technique for forecasting forward interest rates. Its performance is compared to that of a recursive method which has produced some of the best results in previous studies for forecasting forward interest rates.