Crude Oil Consumption Forecasting Using Classical and Machine Learning Methods

The global oil market is the most important of all the world energy markets. Since crude oil is a non-renewable source, its quantity is fixed and limited. To manage the available oil reserves, it will be helpful if we have estimation about the future consumption requirements of this resource beforehand. This paper describes methods to forecast crude oil consumption of next 5 years using past 17 years data (2000-2017). The decision making process comprised of: (1) Preprocessing of dataset, (2) Designing forecasting model, (3) Training model, (4) Testing model on test set, (5) Forecasting results for next 5 years. The proposed methods are divided into two categories: (a) Classical methods, (b) Machine Learning methods. These were applied on global data as well as on three major countries: (a) the USA, (b) China, (c) India. The results showed that the best accuracy was obtained for polynomial regression. An accuracy of 97.8% was obtained.