A Hybrid SARIMA-LSTM Approach for Forecasting Crude Oil Prices

An increase in crude oil price causes some negative effects on oil producers and non-producers. The increase in global commodity prices helps the financial revenue and foreign currency balance of oil exporters. Still, there is a complex situation   for oil-exporting nations, including Nigeria, that sell crude oil and import refined oil. This paper uses a sophisticated hybrid seasonal autoregressive integrated moving average-long short-term memory model to forecast the price of crude oil in Nigeria. The Seasonal autoregressive integrated moving average model, which is one of the traditional statistical models, can handle linear and seasonality issues. On the other hand, long short-term memory model, a deep learning model, effectively deals with non-linear relationships and temporal dependencies. Therefore, this work proposes a hybrid model by combining the advantages of the two models for improuoiving forecasting accuracy. For model selection, two metrics were considered, root mean squared error and mean absolute error. Crude oil prices per month in dollars were obtained, and their range was from January 1946 to June 2025 with 954 data points. The hybrid model demonstrated the lowest mean absolute error (21.66) and root mean square error (21.69) for cross-validation datasets. The study suggests a hybrid model for forecasting crude oil prices in Nigeria.