|
In this study, the author presents a new model to deal with four major issues of fuzzy time series (FTS) forecasting, viz., determination of effective lengths of intervals (i.e., intervals which are used to fuzzify the numerical values), repeated fuzzy sets, trend associated with fuzzy sets, and defuzzification operation. To resolve the problem of determination of length of intervals, this study suggests the application of an artificial neural network (ANN) based algorithm. After generating the intervals, the historical time series data set is fuzzified based on FTS theory. In part of existing FTS models introduced in the literature, each fuzzy set is given equal importance, which is not effective to solve real time problems. Therefore, in this model, it is recommended to assign weights on the fuzzy sets based on their frequency of occurrences. In the FTS modeling approach, fuzzified time series values are further used to establish the fuzzy logical relations (FLRs). To determine the trends associated with the fuzzy sets in the corresponding FLR, this article also introduces three trend-based conditions. To deal repeated fuzzy sets and trend associated with them, this study proposes a new defuzzification technique. The proposed model is verified and validated with real-world time series data sets. Empirical analyzes signify that the proposed model has the robustness to deal one-factor time series data sets very efficiently than existing FTS models. Experimental results show that the proposed model also outperforms over the conventional statistical models.
|
