An evolutionary and connectionist approach for time series forecasting

Abstract

The combination of the evolutionary and connectionist paradigms for problem solving takes a strong inspiration from living systems and is gaining an increasing attention when it comes to the development of computional systems that can handle complex and dynamic problems. One's claim is that Time Series Forecasting is a fertile domain for the test of these technologies. Therefore, a number of experiments were conducted in order to evaluate the merits or demerits of the approach, being the results compared with those obtained from the use of conventional procedures (e.g., the Holt-Winters and the ARIMA ones).