A Fuzzy Time-Series Prediction by GA Based Rough Sets Model

抄録

Fuzzy time-series (FTS) has been applied to handle non-linear problems, such as enrollment, weather and stock index forecasting. In the forecasting processes, fuzzy logical relation (FLR) plays a pivotal role in forecasting accurately. Usually FTS uses an equal interval to obtain forecasted values. However, in this paper, we use genetic algorithms (GA) to optimize the interval at first. Based on this, then rough sets (RS) method is used to recalculate the values. In the empirical analysis, Japan stock index is used as experimental data sets and one fuzzy time-series method, as a comparison model. The experimental results showed that the proposed method is more efficient than the FTS method.