基于特征工程的煤炭价格预测模型研究
Research on coal price prediction model based on feature engineering
杨振乾1
作者信息
- 1. 国能销售集团有限公司,北京 100089
- 折叠
摘要
煤炭价格的精确预测对确保国家能源安全和助力"双碳"目标实现具有重要意义,针对煤价预测准确率仍可进一步提升的问题,选取2017-2023年秦皇岛Q5500动力煤价格及影响因素数据,基于特征工程分析确定煤价预测模型的特征指标分别建立LSTM、RF、SVR和XGBoost模型,根据算法评估指标,比较并分析各模型RMSE、MAE、MAPE、R2性能,结果表明LSTM煤价预测模型的误差更低、准确率更高且精度最优.
Abstract
Accurate prediction of coal price is of great significance to ensure national energy security and help realize the dual-carbon goal,for the problem that the accuracy of coal price prediction can still be further improved,the data of Qinhuangdao Q5500 power coal price and influencing factors in 2017-2023 are selected to determine the feature indexes of coal price prediction model based on the feature engineering analysis,and establish LSTM,RF,SVR and XGBoost model,according to the algorithm evaluation index to compare and analyze the performance of each model RMSE,MAE,MAPE,R2,the results show that the LSTM coal price prediction model has lower error,higher accuracy and optimal precision.
关键词
煤价预测/特征工程/LSTM模型/皮尔森相关系数/机器学习/算法评估指标Key words
coal price forecasts/feature engineering/LSTM model/PPMCC/machine learning/algorithm evaluation index引用本文复制引用
出版年
2024
NETL
NSTL
万方数据