Precision assessment of the machine learning tools for the strength optimization of environmental-friendly lightweight foam concrete

Foamed concrete (FC) is increasingly used in modern construction due to its lightweight nature, superior thermal insulation, and sustainable properties. However, accurately predicting its compressive strength remains a challenge due to the complex interactions of its components. This study addresses this gap by employing advanced machine learning tools, including decision tree (DT), bagging, and AdaBoost, to develop predictive models for FC strength. The results provide a significant improvement in prediction accuracy, offering a reliable tool for optimizing FC design in construction applications. This research aims to streamline the sample creation process in the laboratory, minimize the waiting time for sample testing, and reduce the project's overall cost for researchers. A total of 149 data points were used from the literature to prepare a proper data set for modelling purposes. The modelling procedure used Python code via the Anaconda Navigator software. The statistical evaluation of the metrics, such as R², MAE, and RMSE, along with the sensitivity analysis to check the impact of inputs and the 10-fold cross-validation method to validate the performance, were part of the presented research. Compared to the DT and bagging models, the results demonstrate that the AdaBoost model forecasts FC's compressive strength (CS) more accurately. The AdaBoost model gives the R² value equal to 0.97, while DT and bagging show 0.86 and 0.94, respectively. The lower error result for the AdaBoost model and higher for both DT and bagging indicates the superior precision level of the AdaBoost approach. Finally, the graphical user interface (GUI) was designed utilizing the implemented models, which indicates the additional positive aspect of the presented study.