基于CNN-LSTM-Attention 组合模型的黄金周旅游客流预测——以大理州为例  

Prediction of tourist flow during Golden Week holidays based on CNN-LSTM-Attention combined model: A case study of Dali prefecture, Yunnan Province

作　　者：戢晓峰[1,2] 郭雅诗 陈方 黄志文 李武 JI Xiaofeng;CUO Yashi;CHEN Fang;HUANG Zhiwen;LI Wu(Faculty of Transportation Engineering,Kunming University of Science and Technology,Kunming 650500;Yunnan Integrated Transport Development and Regional Logistics Management Think Tank,Kunming 650500;Faculty of Marxism,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学交通工程学院,昆明650500 [2]云南综合交通发展与区域物流管理智库,昆明650500 [3]昆明理工大学马克思主义学院,昆明650500

出　　处：《干旱区资源与环境》2025年第3期200-208,共9页Journal of Arid Land Resources and Environment

基　　金：国家自然科学基金项目(42061030);云南省交通运输厅科技创新及示范项目(2023-85-2)资助。

摘　　要：黄金周旅游客流预测一直是区域旅游管理的重大现实需求,能够为黄金周旅游组织提供更为精准的数据支持。文中基于百度迁徙数据和百度搜索指数数据,以卷积神经网络(CNN)、长短期记忆网络(LSTM)以及注意力机制(Attention)为基准,构建了CNN-LSTM-Attention组合模型,对大理州黄金周日度旅游客流人数进行了预测,并基于SHAP算法进行了影响因素分析。结果显示:1)CNN-LSTM-Attention组合模型的预测精度优于RF模型、SVM模型、CNN模型、LSTM模型和CNN-LSTM模型。2)引入百度搜索指数特征后,模型的均方根误差(RMSE)、平均绝对百分比误差(MAPE)、决定系数(R^(2))表现最优,表明百度搜索指数的加入在一定程度上提升了模型的预测精度。文中所构模型为黄金周旅游客流预测提供了新思路。Prediction of tourist flow in Golden Weeks has been a major practical need for regional tourism management, which can provide more accurate data support for Golden Week tourism organization. Based on Baidu migration data and Baidu search index, a CNN-LSTM-Attention combination model is constructed with convolutional neural network(CNN), long and short-term memory network(LSTM), and attention mechanism(Attention) as the benchmarks, and the daily tourist flows during the Golden Week holidays in Dali Prefecture are predicted, and the influencing factors are analyzed based on the SHAP algorithm. The results show that: 1) The prediction accuracy of the CNN-LSTM-Attention combined model outperforms that of the RF model, SVM model, CNN model, LSTM model and CNN-LSTM model. 2) After the introducing the the Baidu Search Index features, the model exhibits optimal performance in terms of root mean square error(RMSE), mean absolute percentage error(MAPE), and coefficient of determination(R^(2)). It suggests that the inclusion of the Baidu Search Index has improved the model's prediction accuracy to a certain extent. The model proposed in this paper provides a new idea for predicting tourist flow during the Golden Weeks.

关 键 词：客流预测 黄金周 卷积神经网络(CNN) 长短期记忆网络(LSTM) 注意力机制 

分 类 号：F590[经济管理—旅游管理]