Car Sales Prediction System Based on Fuzzy Time Series and Adaptive Neuro Fuzzy Inference System

This article explains the forecasting algorithm using fuzzy logic. The algorithm involves selecting fuzzy classes to generate forecasting data and selecting membership functions that play an important role in forecasting accuracy. The five main loops in this algorithm involve adaptive nodes and non-adaptive nodes. The membership function included in parameters that only work in adaptive nodes that will increase accuracy along with the number of iterations. The algorithm applied using the Sturges formula to replace iteration behavior in fuzzy logic. The evaluation methods used are Average Forecasting Error Rate (AFER) and Mean Squared Error (MSE). The results showed that the ANFIS algorithm was better with AFER values of less than 15% compared to the fuzzy time series that had errors of more than 20%. The MSE value of ANFIS is also far below the fuzzy time series.