Weight Optimization of The Neural Fuzzy System (NFS) Using Genetic Algorithm for Forecasting

Abstract

Inflation is a phenomenon of increasing prices on a continuous basis which results in the increase of other goods. This study proposes the Neural Fuzzy System (NFS) as a method to predict the rate of inflation in Indonesia. To improve the accuracy, the weight at this stage of Neural Network to be determined correctly. So, this research using Genetic Algorithms to determine the best weights in the training process. This weight can be used to obtained output thorough testing process. Then, it can be processed again in the next step using FIS Sugeno until obtained the end forecasting result. To increase more accurate forecasting results, the establishment of fuzzy rules must be specified correctly. It takes a novelty that minimizes the number of fuzzy rules by dividing the initial parameter into the two (positive and negative) on stage Neural Network. So, the fuzzy rules generated less. To measure the accuracy of the system used the RMSE technique. Based on this result, the proposed method obtained for 0.89.