Abstract
This paper mainly studied the chaotic characteristics and prediction of WTI crude oil monthly price time series from January 1980 to June 2017. Meanwhile, we analyzed whether the major shock of the financial crisis in July 2008 would break the chaotic character of the time series. In addition, when using the largest lyapunov exponent to determine chaotic characteristics, the robustness test of the largest lyapunov exponent was carried out using bootstrap method. Then, we utilized three types of prediction models (ANN+Chaos-type models, Chaos-type model and ANN-type models) to predict the price of crude oil in different months. And we found that the prediction accuracy of ANN-type model is lower than the other type models. This indicated that the accuracy of the prediction with ANN model under the model misspecification is not high because the time series of WTI crude oil price has chaotic characteristics. At last, we constructed a new predictive model, namely HWP-CHAOS model, to compare the prediction accuracy of the above three type models, and discovered the best prediction model among these models is HWP-CHAOS model.
Highlights
Crude oil plays an increasingly important role in the world economy because nearly two-thirds of global energy demand comes from crude oil [1]
The main contributions of this paper are as follows: (1) using bootstrap method to test the robustness of the calculated value of the largest lyapunov exponent; (2) in the field of crude oil, we built up three Artificial Neural Networks (ANN)+Chaos models to predict West Texas Intermediate (WTI) crude oil prices, and in particular, one of them is proposed for the first time, which is called the Hybrid based on Percentage error for the Weight (HWP-CHAOS) model, which provides a new perspective for crude oil price forecast
The prediction ability of the model was measured from Mean Absolute Error (MAE) and percentage error (Perr), and the comparison showed that the prediction ability of the model was ranked from high to low as follows: RBF-CHAOS > RBF, Back Propagation (BP)-CHAOS > BP, which means the accuracy of the prediction model combining chaos with ANN is higher than that of the ANN model
Summary
Crude oil plays an increasingly important role in the world economy because nearly two-thirds of global energy demand comes from crude oil [1]. The main motivation of this study is attempt to propose Chaos-based Artificial Neural Network (ANN) models for WTI crude oil spot price prediction and compare their prediction performance with some other existing forecasting models. The main contributions of this paper are as follows: (1) using bootstrap method to test the robustness of the calculated value of the largest lyapunov exponent; (2) in the field of crude oil, we built up three ANN+Chaos models to predict WTI crude oil prices, and in particular, one of them is proposed for the first time, which is called the Hybrid based on Percentage error for the Weight (HWP-CHAOS) model, which provides a new perspective for crude oil price forecast.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
