Predicting Crude Oil Price Using Time Series Statistical Modelling Techniques

Abstract

One of the important topics in Global Economy is crude oil prediction. Nonetheless, crude oil prediction is related to many factors that play a role in it and include geological factors, production, technical etc…Accurate prediction of crude oil is significant considering all the previous factors. However, these factors are volatile and vary under different conditions. Therefore, this study proposes using machine learning models for time series predictions of crude oil. More specifically, three types of convolutional neural networks were used being autoregressive integrated moving average (ARIMA), seasonal autoregressive moving average (SARIMA) and long short term memory (LSTM). The models were applied to three datasets of crude oil including Crude Oil Daily Price Data, Crude Oil Weekly Inventory Data and Crude Oil Monthly Production Data. Data analysis was conducted using Python. The outcomes of the predictions showed that LSTM had higher accuracy and prediction than ARIMA and SARIMA. The LSTM daily model achieved a MAPE score of as low as 0.087. Future research involve more reliable production data and include more reliable input parameters to improve accuracy.