Generalized Fuzzy Time Series Forecasting Model Enhanced with Particle Swarm Optimization

Abstract

1.IntroductionSince fuzzy time series model was proposed by Song and Chissom [1][2][3] in 1993, there are many forecasting models have been developed to deal with the forecasting problems due to their capability of dealing with the uncertainty and vagueness inherent in the data collected. In the literatures, the experimental analysis existing in many areas, such as forecasting stock price [4][5], tourism demand[6], temperature [7], amount of export [8] and dry bulk shipping index [9], etc.Most of the exiting methods mainly focused on partitioning the universe of discourse, constructing fuzzy relationships from the fuzzy set, forecasting and defuzzifing the forecasting output. A proper choice of the length of each interval can greatly improve the forecasting results. Huarng [10,11] presented some methods to obtain the proper lengths of the intervals for the fuzzy time series model. Chen proposed the automatic clustering technique to generate clustering-based intervals for some fuzzy time series forecasting models [12,13].Literatures [14,15]introduced an approach which using an optimization technique with a single-variable constraint to determine an optimal interval length. The particle swarm optimization algorithm and some other algorithms were used to search the appropriate intervals on the fuzzy time series [16][17][18][19] as well geneticalgorithms [20]. To improve the fuzzy time series model in constructing fuzzy relationships, forecasting and

A Comparative Study and Analysis of Time Series Forecasting Techniques

“…In [14], a weighted fuzzy model has been applied on Taiwan stock index data that proved weighted model performed better than conventional fuzzy time series models. In [15,16], particle swarm optimization has been introduced with fuzzy NN to optimally find the intervals which increases the forecasting accuracy.…”