Price Prediction of ARIMA-ES-RF Brent Crude Oil Based on ACO Optimization
To address the issue of significant errors in forecasting Brent crude oil prices using conventional single-method approaches like autoregressive integrated moving average(ARIMA)and exponential smoothing(ES),which struggle to precisely capture the nonlinear characteristics of the series,a combined forecasting method is proposed:autoregressive integrated moving average(ARIMA)-exponential smoothing(ES)-random forest(RF).While numerous models for predicting oil prices exist,no literature has yet explored the use of random forest in combination with traditional time series models for oil price forecasting.Building upon this,ant-colony optimization(ACO)is utilized to intelligently search for critical parameters(number of trees and depth of roots)within the Random Forest.The predictive accuracy of the combined model with Random Forest is fed back to the ant colony for real-time updates of information pheromones,yielding optimal model parameters that maximize predictive accuracy.The research findings demonstrate that the combined random forest model,optimized with ant colony parameters,yields improved trend prediction for Brent crude oil prices.The root mean square error(RMSE)drops from 1.15 to 0.88,marking a reduction of 0.27.Additionally,the mean absolute percentage error(MAPE)decreases from 1.10％to 0.86％,a reduction of 0.24％.This enhancement in predictive accuracy significantly surpasses previous models used for oil price forecasting.
random forestant ccolony algorithmBrent crude oil price
万方数据
维普
