Meta fuzzy functions: Application of recurrent type-1 fuzzy functions

Abstract

The main objective of meta-analysis is to aggregate the results of multiple scientific studies on a specific topic. Instead of aggregating the results of different studies, different methods are aggregated with the help of fuzzy c-means clustering algorithm in the proposed method. Meta fuzzy functions are introduced in the paper. The idea of meta fuzzy functions is to aggregate the methods which are proposed for the same purpose; forecasting, prediction, etc. The study aggregates the models for the same method under different parameter specifications rather than aggregating different methods. Recently, recurrent type-1 fuzzy functions are introduced as an alternative forecasting method. The main advantages of recurrent type-1 fuzzy functions are that they are free of assumptions and rules. There are three parameters to be adjusted for recurrent type-1 fuzzy functions; the number of lags for AR(p), the number of lags for MA(q), and the number of clusters. The models for recurrent type-1 fuzzy functions with different parameter specifications are aggregated in the paper. The results show that it is possible to increase the forecasting performances of recurrent type-1 fuzzy functions in terms of both RMSE and MAPE.