Restaurant Customer Flow Prediction Based on CNN-LSTM-Attention Model

Customer flow prediction is essential for optimizing resource allocation and enhancing operational efficiency in restaurants. This paper addresses the challenge of predicting restaurant customer flow by incorporating external data, such as platform views and weather conditions, alongside feature derivation and selection to improve data quality. For the model design, we propose a hybrid prediction model based on CNN-LSTM-Attention. The CNN is used for local feature extraction, the LSTM captures the long-term dependencies of the time series, and the Attention mechanism dynamically focuses on key features. Through comparative and ablation experiments, we demonstrate that the proposed model significantly outperforms other benchmark models in metrics such as MAPE, RMSE, and R², indicating superior prediction performance. Additionally, SHAP-based explainability analysis further illuminates the influence of key features on the prediction results, enhancing the model's explainability and practical application value.