In steel plate girders (SPGs), a patch loading usually causes a local failure in the vicinity of the loading area of the girder web. However, estimating the patch loading resistance (PLR) of SPGs is challenging due to the complexity of the problem. This paper aims to develop a novel hybrid WOA-GBM model based on a whale optimization algorithm (WOA) and a gradient boosting machine (GBM) for predicting the PLR of longitudinally SPGs. Firstly, 137 tests of longitudinally stiffened SPGs subjected to patch loading are carefully collected and divided into training and test sets. Then, the most critical parameters of the GBM model are determined using 10-fold cross-validation integrated with the WOA. The results obtained from the WOA-GBM model are compared with those from adaptive boosting (AdaBoost) and extreme gradient boosting (XGBoost) models. The results show that the WOA-GBM model outperforms other models. Additionally, SHapley Additive exPlanation (SHAP) method is used to explain the prediction of the proposed WOA-GBM model globally and locally. Finally, an efficient graphical user interface (GUI) tool and a web application (WA) are developed to apply the proposed WOA-GBM model for practical use.