Tracking federal funds target rate movements using artificial neural networks

Abstract

The Federal Reserve determines federal funds target rate (FFTR), which is one of the most publicized and anticipated economic indicator in the financial world. As the decision making process is complex due to unknown functions, it has been a difficult and challenging process for the researcher to model the thoughts of the Federal Open Market Committee (FOMC) members using statistical methods and hence predict the changes in FFTR. With artificial neural networks evolving as an efficient and promising methodology, it is possible to emulate the decision making of FOMC. In this paper, two-level neural network architecture has been established to forecast the direction and magnitude of changes in FFTR. First level is the self-organizing map (SOM) and the second level is the general regression neural network (GRNN). The period of study is during the term of Chairman Alan Greenspan, where the Fed emphasizes largely on the economic time series to make their decision. This paper aims to investigate the effect of one-level neural network, which is GRNN and two-level neural network, which is SOM and GRNN on the prediction of direction and magnitude of changes of FFTR. The result of the comparison is presented in this paper.