【摘要】 基于石油价格易受政治、经济、文化等多种因素的干扰,借助遗传算法和支持向量机对单一的ARIMA模型进行改进,提出GA-SVM-ARIMA模型对石油价格进行预测。结果表明,该预测模型通过了平稳性和白噪声检验,学习率和训练损失值明显优于ARIMA模型和SVM-ARIMA模型。在应用效果中,GA-SVM-ARIMA模型对石油价格的预测精度较高,与真实值之间的误差不超过2%,不仅在短期价格预测中的平均绝对误差均值(2.162 5%)和均方根误差值(3.227%)明显低于其他模型,而且在长期价格预测中也表现出较高的精准度。更多还原

【Abstract】 Using genetic algorithms and support vector machines to improve a single ARIMA model and proposing the GA-SVM-ARIMA model to predict oil prices based on various factors such as politics,economy and culture.The results showed that the predicted model passes the stability and white noise inspection,and the learning rate and training loss are significantly better than the ARIMA model and SVM-ARIMA model.In the application effect,the GA-SVM-ARIMA model has a high accuracy of oil prices,and the error between the real values does not exceed 2%.This model not only has significantly lower mean absolute error(2.162 5%) and root mean square error(3.227%) in short-term price prediction than other models,but also shows a high accuracy in long-term price prediction.更多还原