Artificial Intelligence in Crude Oil Price Forecasting

Abstract

The crude oil is a common energy source for nearly all commercial sectors, its price forecasting activities have always been an important issue for both governments and commercial firms to make better decisions and investments. In this research, both the history of the crude oil price forecasting and used artificial intelligence methods on forecasting were investigated. In early stages of crude oil price forecasting, traditional statistical and mathematical models were used, while afterwards computer based artificial intelligence models became more popular. These models were more appropriate to the non-linear, volatile and complex structure of oil prices. Artificial intelligence gave chance to evaluate the situation in many aspects at the same time with the help of the computers’ power. Evaluation of these produced outcomes together with other variables such as historical prices, weather condition, political situations etc. gave much better forecasting results for crude oil. Fuzzy logic, Artificial Neural Network (ANN), Genetic Algorithms (GA), Support Vector Machine (SVM), expert systems, text-mining algorithms and their sub-versions were the frequently used AI based algorithms in the crude oil forecasting models. Among them for forecasting of crude oil prices, ANN algorithms, with its layered structure which makes it possible to relate many parameters with target variable a detailed way, have the most appropriate working principle for forecasting the complex and sensitive structure of crude oil prices. Hybrid models usually give better results, its combination with other algorithms such as text-mining or the most used one ANN, could improve the prediction results.