Analysis of membership function in implementation of adaptive neuro fuzzy inference system (ANFIS) method for inflation prediction

Abstract

This research will analyze which fuzzy membership function (MF) gives the best results in the implementation of the Adaptive Neuro-fuzzy inference system (ANFIS) method. The case study conducted is to predict the growth of inflation in Bali Province with ANFIS which has the main objective of analyzing the fuzzy membership function and designing a model that can predict the value of inflation growth. Inflation can also be defined as a process of increasing general prices or decreasing the value of money continuously. Inflation growth prediction uses the ANFIS method with five input parameters in the form of regional economic indicators, and the number of pairs of initial data used is 34 annual periods. Several types of membership functions (MF) that will be tested and analyzed are triangular MF, MF trapezium, and MF gbell. The cryptic inference system used is TSK-Order One, and the learning method used is a hybrid method. Based on the research results, the analysis of the fuzzy membership function in the inflation prediction system produces the best error is 1.35E-07 with the type of triangular membership function (MF).