The Effectiveness of Hybrid Backpropagation Neural Network Model and TSK Fuzzy Inference System for Inflation Forecasting

Abstract

Forecasting may predict the accurate future condition based on the previous circumstance. Problems that may occur are related to forecasting accuracy. This study proposes a combination of two methods: Neural Network (NN) and Fuzzy Inference System (FIS) to accuratelly forecast the inflation rate in Indonesia. Historical data and four external factors were used as system parameters. The external factors in this study were divided into two fuzzy sets. While time series variables were divided into three fuzzy sets. The combination of them generated a lot of fuzzy rules that may reduce the forecasting effectiveness. As a consequence, the less fit fuzzy rules formation would produce a low accuracy. Therefore, grouping all input variables into positive parameters and negative parameters are necessary for efficiency improvement. To evaluate the forecasting results, Root Means Square Error (RMSE) analytical technique was used. Fuzzy Inference System Sugeno was used as the base line. The results showed that the combination of the proposed method has better performance (RMSE=2.154901) than its base line.