摘要
从研究煤炭价格序列自身变化规律的角度,提出基于滚动时间窗的ε-SVR预测模型,通过数据重构获得输入与输出样本,随着时间推移,不断更新滚动时间窗的数据内容,从而建立动态的ε-SVR模型预测最新时点的煤炭价格。将此模型应用于秦皇岛港5500 kcal混煤价格的预测,分别进行了1期、3期、6期、9期及12期的价格预测,所有预测结果的平均误差值不超过3%,可见预测精度较高,预测效果良好。而且此模型数据获取简单、计算灵活方便,可应用于煤炭价格等非平稳时间序列的预测问题中,其结果可以为相关企业决策者提供科学有效的数据支持。
From the perspective of studying the regular pattern of changes in coal price series,a prediction model using method is proposed based on the sliding time window.In this method,input and output samples are obtained by data reconstruction,As time goes on,the data content of the sliding time window is constantly updated,thereby establishing a dynamicε-SVR model to predict coal prices at the latest time.This model was applied to the prediction of the 5500 kcal mixed coal price in Qinhuangdao Port,and the 1-term,3-term,6-term,9-term and 12-term predictions were carried out respectively,the average error value of all the prediction results did not exceed 3%,which shows that the prediction accuracy is high and the prediction effect is good.Moreover,the data of this model is simple to obtain,flexible and convenient to calculate,so which can be applied to the prediction problems of non-stationary time series such as coal prices,the results can provide scientific and effective data support for related enterprise decision makers.
作者
宁晖
周文文
Ning Hui;Zhou Wenwen(China Coal Energy Research Institute Co.,Ltd.,Xi’an 710054,China;School of Economics and Management,Beijing University of Technology,Beijing 100124,China)
出处
《煤炭经济研究》
2020年第3期12-18,共7页
Coal Economic Research
