Time Series Forecasting Model Based on Intuitionistic Fuzzy Set via Equal Distribution of Hesitancy De-I-Fuzzification

Abstract

The fuzzy time series forecasting model is a powerful tool in forecasting the time series data. The nature of the fuzzy set exhibits its role in handling the uncertainty of the data. The intuitionistic fuzzy set (IFS) is a generalization of a fuzzy set that makes the forecasting process more precise and accurate. This paper proposes a new fuzzy forecasting model based on IFS via the de-i-fuzzification approach, namely equal distribution of hesitancy. The proposed model consists of four main parts; the fuzzification of historical data; the establishment of the IFS; the de-i-fuzzification; and the defuzzification. For the fuzzification, the historical data is partitioned into 14 intervals using the frequency density-based method and trapezoidal fuzzy numbers are used to fuzzify the data. The data are then converted into IFS. The data in IFS form is reduced to fuzzy set using equally distributed with the degree of hesitancy approach. The arithmetic rules based on centroid defuzzification is used to calculate the forecasted output. The proposed model shows a better performance than the existing forecasting models based on IFS, indicating that the equal distribution of hesitancy de-i-fuzzification managed to handle the non-determinism in the forecasting with simplified procedure. In the future, an improved method will be proposed to defuzzify the IFS into crisp values without going through the de-i-fuzzification process, yet preserving the nature of IFS.