Biologically Inspired Fuzzy Forecasting: A New Forecasting Methodology

Abstract

There are many forecasting techniques including the ARIMA model, GARCH model, exponential smoothing, neural networks, genetic algorithm, etc. Those methods, however, have their drawbacks and advantages. Since financial time series may be influ- enced by many factors, such as trading volume, business cycle, oil price, and seasonal factor, conventional model based on prediction methodologies and hard computing meth- ods seem inadequate. In recent years, the innovation and improvement of forecasting methodologies have caught more attention, and also provide indispensable information in the decision-making process, especially in the fields of financial economics and engi- neering management. In this paper, a new forecasting methodology inspired by natural selection is developed. The new forecasting methodology may be of use to a nonlinear time series forecasting. The method combines mathematical, computational, and biological sciences, which includes fuzzy logic, DNA encoding, polymerase chain reaction, and DNA quantification. In the empirical study, currency exchange rate forecasting is demonstrated. The Mean Absolute Forecasting Accuracy method is defined for evaluating the performance, and the result comparing with the ARIMA method is illustrated.