The article presents the research of the application of machine learning (ML) technologies for determining the strength of the walls of reinforced concrete I-beams. The current experience of using ML methods, such as neural networks, random forest and gradient boosting, is analyzed, the prospects of hybrid methods for predicting the degree of dependence and other strength characteristics of reinforced concrete structures are discussed. A hybrid model predicting the strength of the walls was developed based on the CatBoost algorithms and implemented in the Jupyter Notebook environment in Python. The model was trained on experimental data containing various parameters, designs and properties of materials, including the geometric dimensions of the beams, types of reinforcement and concrete class. Statistical analysis of the experimental sample for each instance was carried out. The regularization method was used to normalize the experimental data, optuna was used for their optimization, gridsearch, random search was used for the selection of hyperparameters. The data was checked for multicollinearity of features. Histograms of joint dependencies of variables were constructed. Practical and predicted values of the model were visualized in the form of a scatter plot. The most significant parameters affecting the strength of the walls of reinforced concrete I-beams were obtained. The constructed model passed the quality assessment according to three metrics: MAE = 0.2138; MSE = 0.4885; MAPE = 0.0033.