Neural network-based fuzzy auto-regressive models of different orders to forecast Taiwan stock index

This article proposes the use of a fuzzy time series model based on neural networks that are intended to calculate the complicated fuzzy relationships among observations. The Taiwan stock exchange capitalisation weighted stock index is used as the forecasting target. Various parameters such as the order of the time series the threshold for defuzzification, and the in-sample estimation results are used to determine the proper models for out-of-sample forecasting.