Based on the principal component analysis(PCA)method, the African vulture optimization algorithm(AVOA), and lightweight gradient learning machine(LightGBM)model, a stress prediction model of concrete dam based on PCA-AVOA-LightGBM was constructed.PCA method was used to mine the main influencing factors of dimension reduction-based stress prediction, and AVOA was introduced to optimize the hyperparameters of the LightGBM model.Based on the stress monitoring data of a concrete dam, the PCA method was applied to vector regression machine (SVR), random forest (RF), extreme gradient boosting (XGboost), LightGBM models, and a comparative analysis was conducted with the PCA-AVOA-LightGBM model.The results show that the PCA method effectively reduces the multicollinearity among the influencing factors of each model. The PCA-AVOA-LightGBM model shows better performance in accuracy and efficiency than other prediction models and can be applied in stress monitoring of similar concrete dams.