[Purposes] Vehicle load is one of the critical factors causing deflection variation of bridges during the service stage, characterized by strong spatial-temporal randomness and great challenges in prediction. To improve the prediction accuracy of deflection induced by vehicle loads, this paper establishes a deep learning model of CNN-LSTM-Attention integrated with feature engineering, and systematically analyzes the effects of feature engineering and model structure on prediction performance. [Methods] Firstly, based on data from the bridge deck monitoring and weigh-in-motion systems, a lane-level vehicle load time series is constructed. Feature engineering methods including lag features, difference features, rolling standard deviation, and moving average are introduced to enhance the representation capability of the time series. On this basis, the CNN-LSTM-Attention prediction model is established and compared with the traditional LSTM model and the model without feature engineering. Finally, the model performance is quantitatively evaluated using indicators such as MAE, MSE, R2, and explained variance score (EVS). [Findings] The results indicate that the prediction accuracy of the model is significantly improved after introducing feature engineering, with a MAE of 0.8235, MSE of 1.4174, R2 of 0.9962, and EVS of 0.9962. Compared with the model without feature engineering, the MAE is reduced by 87.2% and R2 is increased by 27.7%. Compared with the traditional LSTM model, the MAE is reduced by 3.75% and R2 is improved by approximately 34.5%. [Conclusions] The CNN-LSTM-Attention model integrated with feature engineering can effectively characterize the temporal features and long?term dependencies of vehicle load series, thereby achieving high?precision prediction of bridge deflection. Optimizing only the network structure provides limited performance improvement, whereas the combined strategy of “feature engineering + deep learning model” serves as a key approach to enhancing the prediction accuracy of vehicle?induced deflection. This work provides a novel data?driven technical tool for bridge structural health monitoring and operation maintenance management.