International crude oil price forecasting with news influence index attenuation

With the development of big data technology, news data has been applied to forecast international crude oil price, but there is a lack of research on the impact degree and duration of news data. In order to quantify the decay of news influence, this paper puts forward a new exponential decay analysis method based on exponential decay and mutual information. In order to improve the prediction accuracy, this paper has constructed a combined kernel function support vector regression prediction model based on differential evolution optimization algorithm which realizes the optimal selection of weight coefficient, kernel function parameters and regression model parameters. To evaluate the effectiveness of the proposed methods, eight models are selected for comparative study. The empirical results show that integration of news influence decay time series can improve the correlation and prediction accuracy of crude oil price; the proposed forecasting method has good prediction accuracy, the average absolute percentage error of verification set is 1.53%, which is better than the comparison models. © 2022 Systems Engineering Society of China. All rights reserved.