摘要
针对兰州空气质量指数存在波动大和数据长期依赖性的问题,提出了一种基于CEEMDANLSTM组合的预测模型,并与EEMD-LSTM和LSTM模型进行了比较。首先采用CEEMDAN对兰州空气质量指数序列进行分解,然后使用LSTM神经网络预测得到各个分量,最后叠加各分量的预测值重构空气质量指数预测结果。实验结果表明,CEEMDAN-LSTM模型相比于LSTM模型和EEMD-LSTM模型,具有更小的预测误差和更高的预测精度。这得益于CEEMDAN方法的有效降噪和LSTM模型对长期依赖关系的强大处理能力。因此,该组合模型在兰州空气质量指数预测方面具有一定的实用价值。
Aiming at the problems of large fluctuation and long-term data dependence of Lanzhou AQI,a prediction model based on CEEMDAN-LSTM is proposed in this paper,and compared with EEMD-LSTM and LSTM models.Firstly,CEEMDAN was used to decompose the AQI sequence of Lanzhou,then LSTM neural network was used to predict each component,and finally the predicted value of each component was added to reconstruct the AQI prediction result.The experimental results show that CEEMDAN-LSTM model has smaller prediction error and higher prediction accuracy than LSTM model and EEMD-LSTM model.This is due to the effective noise reduction of CEEMDAN method and the strong handling ability of LSTM model for long-term dependencies.Therefore,the combined model has certain practical value in Lanzhou AQI prediction.
作者
赵煜
韩旭昊
ZHAO Yu;HAN Xu-hao(School of Statistics,Lanzhou University of Finance and Economics,Lanzhou 730020,China)
出处
《安徽师范大学学报（自然科学版）》
2023年第5期433-439,447,共8页
Journal of Anhui Normal University(Natural Science)
基金
国家社会科学基金项目(21XTJ004).
