Comparison of ANFIS and NFS on inflation rate forecasting

Abstract

Inflation is very influential on the national economy. The monetary crisis can occur if inflation is not well controlled. For that, it takes a forecasting. Inflation rate forecasting can predict future country situation based on historical data. This study proposes two hybrid Fuzzy logic-Neural network methods to predict inflation rate in Indonesia. Adaptive Neuro Fuzzy Inference System (ANFIS) and Neural Fuzzy System (NFS) were chosen because both methods are hybrid Fuzzy logic-Neural network with different architecture. This study aims to find the method that has the best performance. Time series data and some external factors (CPI, Money Supply, BI Rate, Exchange Rate) are used as parameters. Proper Neural Network (NN) architecture must be found to produce high accuracy. Therefore, some tests (learning rate, epoch, neuron) are performed. The best method is chosen based on the level of accuracy produced by using Root Mean Square Error analysis technique. The results show that NFS has better performance with accuracy (RMSE=1.213) than ANFIS.