A new method for macroeconomic prediction

Macroeconomics is a complex and nonlinear macrosystem, in which various elements interact with each other. Traditional methods to predict and analyze macroeconomics, such as time series, cannot get satisfactory results. Simply applying artificial neural networks (ANN) also cannot achieve good results. This paper introduces a new method, called Trend Fitting ANN (TFANN), which combines polynomial fit with the traditional ANN method. TFANN disassembles historical economic data to the trending and the fluctuating element, and then analyzes the two elements respectively. The experimental results show the algorithm is reliable and efficient.