Comparison of fuzzy time series forecasting model based on similarity measure concept with different types of interval length.

Fuzzy time series forecasting model has been proposed to cater for data in linguistic values. One of the crucial factors that influence the performance of fuzzy time series is the partition of interval length. This paper compares the effect of several interval lengths to the performance of fuzzy time series forecasting model. The interval length considered are the average based, frequency density based and randomly chosen length methods. The data are represented in trapezoidal fuzzy numbers and the accuracy of the forecasting model is calculated using the distance, area, height and perimeter ratio similarity measure. The model is applied in a numerical example of Malaysian unemployment rate. The findings show that the average based length outperforms the other two types of interval length. [ABSTRACT FROM AUTHOR]