Skip to Main Content
International crude oil prices are very complex nonlinear time series, which are not only affected by the domination of objective economic laws, but also by politics and other factors. Therefore it is difficult to establish an effective prediction model based on the general time series analysis. In this paper, based on wavelet transform, the international oil prices time series is decomposed into approximate components and random components. The approximate components, which represented the trend of oil price, are predicted with Boltzmann neural network; the random components are predicted with Gaussian kernel density estimation model. In this paper, we analyzed the time-frequency structure of dubieties wavelet transform coefficient modulus for crude oil price time series, and predicted the oil price with Boltzmann neural network and Gaussian kernel density estimation model.The results show that the model has higher prediction accuracy.