Towards Precision: A Comparative Analysis of Crude Oil Price Forecasting Approaches

This research focuses on the timeseries data forecasting efficiency models such as decision tree, random forest, support vector and multi-linear regressions, and multi-layer perceptron when supplemented with technical indicators such as exponential moving average and moving average convergence divergence. For the study, 3 datasets of crude oil based on time interval/frequency of prediction are taken into action. This includes one day, one week, and one month of records ranging over a 10-year time period starting from 2013 till 2023. The addition of technical indicators as input features increases the total number of features available in the dataset. These indicators are functions of the closing price, the expected outcome which helps greatly in capturing the trends of the price along with the basic features. By use of different evaluation metrics, the models are holistically compared to each other in a try to seek out the best performing model which can be used in real application when predicting price of the crude oil. Across all three datasets, decision tree regression continuously maintained a high degree of accuracy (greater than 0.9). Other models, such as multiple linear regression and the multi-layer perceptron, worked well with datasets that had more records.