Intervals in Fuzzy Time Series Model Preliminary Investigation for Composite Index Forecasting

Many forecasting models have been proposed to improve forecasting accuracy. Recently, Chen at al. model which incorporates three concepts of Fibonacci sequence, framework of Song and Chissom's model and weighted method of Yu's model has been proposed as a method to improve forecasting accuracy. However, the issue on lengths of intervals has not been investigated by Chen et al. despite Huarng advocated that length of intervals affects forecasting results. The issue of effective intervals in determining efficient forecasting has not been conclusively defined. Therefore this paper tests sixteen intervals using randomly chosen length of interval partitioning into the Chen et al. model to investigate this issue. A two-year weekly period of Kuala Lumpur Composite Index data sets were used to demonstrate the effectiveness of sixteen intervals. Empirical results show that that the sixteen intervals using randomly chosen partitioning method can be applied to improve fuzzy time series forecasting.