摘要
为了进一步提高我国货运量的预测准确性,文章基于卷积神经网络和长短期记忆网络模型,引入注意力机制(Attention Mechanism)的组合预测模型,以对我国货运量进行时序预测。首先,利用卷积神经网络提取货运量数据变化特征。其次,将所提取的特征构成时间序列作为长短期记忆网络的输入。最后,通过注意力集中捕捉预测模型中经LSTM层输出的信息特征,划分权重比例,提取关键信息,实现货运量预测。结合全国月度货运量历史数据进行时序预测,然后与其他神经网络预测的各种评价指标进行对比,结果显示,CNN-LSTM-Attention模型预测误差小于其他模型,预测准确性相对较好。
In order to further improve the prediction accuracy of China's high freight volume,this paper introduces a combined prediction model of Attention Mechanism based on convolutional neural network and long and short-term memory network model to forecast China's freight volume in time series.First of all,the convolutional neural network is used to extract the features of the freight volume data changes,and then the extracted features are used to constitute a time series as the input of the long and short-term memory network,and finally,the attention is focused on capturing the features of the information output from the LSTM layer in the prediction model,dividing the weight ratio,extracting the key information,and realizing the prediction of the freight volume.Combined with the national monthly freight volume historical data for time series prediction,and then compared with other neural network prediction of various evaluation indexes,the results show that the CNN-LSTM-Attention model prediction error is smaller than other models,and the prediction accuracy is relatively good.
作者
燕学博
曹世鑫
YAN Xuebo;CAO Shixin(School of Management,Fujian University of Technology,Fuzhou 350118,China;School of Transportation,Fujian University of Technology,Fuzhou 350118,China)
出处
《物流科技》
2024年第14期5-9,共5页
Logistics Sci-Tech
基金
数字经济赋能福建省乡村振兴研究(GY-Z23160)。
