基于Informer神经网络的福建省货运量预测
燕学博1,2，钟凯彬3
1.福建理工大学管理学院，福建 福州 350118；
2.景区交易数据要素化文化和旅游部技术创新中心，福建 福州 350000；
3.福建理工大学互联网经贸学院，福建 福州 350014
摘要：为准确预测货运量，以福建省货运量为研究对象，分别从产业结构水平、经济发展水平、物流发展水平和人文因素等4个方面选取货运量的主要影响因素，以1978—2022年福建省货运量及货运量影响因素数据构建数据集，基于Informer神经网络构建货运量预测模型，通过交叉验证法训练模型，同时采用长短时记忆（long short-term memory, LSTM）神经网络和Transformer神经网络预测福建省货运量，对比三者预测结果的准确性。结果表明：Informer神经网络模型测试集预测结果的平均相对误差为3.75%，LSTM和Transformer神经网络模型预测结果的平均相对误差略高，分别为4.45%、4.38%，Informer神经网络模型的预测结果较准确。采用Informer神经网络模型预测2023年福建省的货运量为184 289.0万t，比2022年增大8.9%。福建省的货运量逐年增大，应不断完善物流配送机制，提高运输效率以满足福建省及周边地区的物流需求。
关键词：货运量；预测；Informer神经网络；交叉验证法
Freight volume prediction of Fujian Province based on Informer neural network
YAN Xuebo1,2, ZHONG Kaibin3
1.School of Management, Fujian University of Technology,Fuzhou 350118, China;
2.Technology Innovation Center of Factored Data in Smart Culture and Tourism, Ministry of Culture and Tourism, Fuzhou 350000, China;
3.School of Internet Economics and Business, Fujian University of Technology, Fuzhou 350014, China
Abstract: To accurately predict freight volume, this study examines the freight volume in Fujian Province by selecting the main influencing factors of freight volume from four aspects: industrial structure level, economic development level, logistics development level, and human factors. A dataset is constructed using data on freight volume and its influencing factors in Fujian Province from 1978 to 2022. Based on the Informer neural network, a freight volume prediction model is constructed and trained using cross-validation methods. Simultaneously, the study employs long short-term memory (LSTM) neural networks and Transformer neural networks to predict the freight volume in Fujian Province, comparing the prediction accuracies of the three models. The results show that the average percentage error of the test set for the Informer neural network model is 3.75%, which is smaller than that of the LSTM and Transformer neural network models, at 4.45% and 4.38%, respectively, which indicates that the Informer neural network model provides more accurate predictions. The Informer neural network model predicts that the freight volume in Fujian Province in 2023 will be 1 842.89 million tons, an increase of 8.9% over 2022. The freight volume in Fujian Province is increasing year by year, and there should be continuous improvements to the logistics distribution mechanism and enhancements in transportation efficiency to meet the logistics demands of Fujian Province and the surrounding regions.
Keywords: freight volume; prediction; Informer neural network; cross-validation
