MIMO evolving functional fuzzy models for interest rate forecasting

Abstract

Forecasting the term structure of interest rates plays a crucial role in portfolio management, household finance decisions, business investment planning, and policy formulation. This paper proposes the use of evolving fuzzy inference systems for interest rate forecasting in the US and Brazilian markets. Evolving models provide a high level of system adaptation and learns the system dynamic continuously, which is essential for uncertain environments as fixed income markets. Besides the usefulness evaluation of evolving methods to forecast yields, this paper suggests the interest rate factors forecasting taking into account multi-input-multi-output (MIMO) evolving systems, which reduces computational time complexity and provides more accurate forecasts. Results based on mean squared forecast errors showed that MIMO evolving methods perform better than traditional benchmark for short and long-term maturities, for both fixed income markets evaluated.