CatBoost algorithms to predict the load-bearing capacity of centrally compressed short CFST columns of circular cross-section

The article investigates the use of regression models, GradientBoostingRegressor and RandomForestRegressor, to predict the load-bearing capacity of centrally compressed short concrete filled steel tubular columns. The work is based on experimental data covering a wide range of column geometric characteristics and material strength characteristics. An important part of the analysis was to deter-mine the influence of various parameters on the model predictions. The importance of features, assessment of the quality of models (MSE, MAE, MAPE) were considered , and visualization of actual and predicted values was carried out to compare the results. The results showed that both models success-fully cope with the task of predicting the load-bearing capacity of structures under given conditions. Analysis of the importance of features revealed the most significant parameters affecting the load-bearing capacity of columns. Visualization of forecasts and analysis of residuals confirmed the adequacy of the models. Additionally, a process of tuning model parameters using cross-validation was carried out to optimize their performance. The results of the study can be used in engineering applications such as the design of reinforced concrete structures to predict load-bearing capacity.