关键词: 交通运输工程；集装箱吞吐量预测；混合注意力机制；多变量输入；CNN-BiLSTM预测模型

Abstract: In order to predict port container throughput more accurately, a prediction model that integrated convolutional neural networks (CNN) with bidirectional long short-term memory (BiLSTM) networks was proposed, and multiple attention mechanisms were also introduced to comprehensively capture the global features of the data. In the proposed model, the influencing indicators and historical container throughput data were combined as multivariate inputs for prediction. The results indicate that, compared to traditional LSTM prediction models and CNN-LSTM hybrid models, both the mean absolute percentage error (MAPE) and root mean square error (RMSE) of the proposed model are reduced, and the model fitting degree (R2) is significantly improved. Notably, in instances of significant data fluctuations, the proposed model achieves more accurate prediction results, which helps port and shipping enterprises adjust their planning decisions and business strategies in a timely manner.

Key words: traffic and transportation engineering; container throughput prediction; hybrid attention mechanism; multivariate input; CNN-BiLSTM prediction model