Evaluation of the Adaptive Fuzzy Neuro Inference System and Fuzzy Model Time Series Markov Chains in Forecasting Crude Oil Prices

The development of a country's economy is greatly influenced by global economic conditions, given the increasingly close links between countries through economic relations and international cooperation. One of the main factors in economic growth is international trade, particularly export and import activities. Crude oil is one of the most actively traded commodities. Given the highly volatile crude oil market, accurate price forecasts are crucial in economic and financial decision-making. This study compares the performance of Adaptive Neuro-Fuzzy Inference System (ANFIS) and Fuzzy Time Series Markov Chain (FTSMC) in forecasting the price of West Texas Intermediate (WTI) crude oil using time series data from 2020 to 2024 with saturated sampling technique. The implementation of both methods is carried out through Matlab Online and R-Studio software, with results showing that ANFIS has higher accuracy than FTSMC, as evidenced by the Mean Absolute Percentage Error (MAPE) value of 1,8010% for ANFIS and 3,7567% for FTSMC. Further analysis shows that ANFIS with a triangular membership function as well as significant lags at lag 1, lag 3, lag 4, and lag 7 is able to produce more accurate predictions and match the trend of actual data. Therefore, ANFIS is recommended as a more effective method in forecasting WTI crude oil prices, which can provide valuable insights for policy makers and industry stakeholders.