Predictive analytics model for forecasting crude oil prices

Abstract

Crude oil is the most important source of energy in the world, accounting for 33% of world energy needs, in comparison to coal and natural gas, which account for 28% and 23%respectively. While liquid fuel consumption is decreasing in comparison to other sources, it is still expected to grow at a rate of 1.2% percent per annum till 2030 (EIA projections).In the last decade, emerging economies such as China and India have progressively increased their share in the consumption of crude oil and non-OECD demand for oil is now at par with demand from OECD countries. In wake of this increased demand and limited supply and reserves, prices are likely to increase. With recent economic shocks and increased speculation in the world commodity markets, volatility is likely to be higher than before. Given the importance of crude oil in the world economy and the inherent risks arising out of high volatility, it is extremely important that forecasting models be built to predict future price movements. We have explored three different approaches to modeling the prices. The first approach consists of stochastic models which do not use historical data. Instead, they make assumptions about the probability distributions of the relevant random variables and attempt to make predictions using the same. The second approach is linear regression. However, this approach is not robust enough to handle the non-linearity in prices. The third and final approach is a model based on fundamental economic/geopolitical factors that uses Artificial Neural Networks. We found this to be the best predictor of both monthly and daily crude oil prices. An important aspect of this study is the increase in prediction accuracy that can result from incorporating market data into our models. We found that the moving average price for the last two days is a good predictor of future daily prices. Factors other than basic demand and supply, such as the behavior of OPEC countries behaving as a cartel (fixing production quotas to control prices), speculation in the world commodity markets, the presence of strategic oil reserves in OECD countries and its absence in emerging economies, severe negative sentiment following the global financial crisis, are all factors that add to the complexity of an already complex and dynamic market environment.