Investigating the compressive property of foamcrete and analyzing the feature interaction using modeling approaches

• Machine learning techniques were used to investigate the compressive strength of foamcrete. • Multi-layer perceptron (MLP) and random forest (RF) machine learning techniques were used for modeling. • SHAP analysis was performed for the feature interaction assessment. • Based on R 2 and errors assessment, RF predictions were more accurate than MLP. • The study's findings are of great importance in eliminating costly laboratory tests. This study aims to investigate the compressive strength (CS) of foamcrete (FC), with a specific emphasis on its 7-day and 28-day performance. The multi-layer perceptron (MLP) and random forest (RF) machine learning techniques have been selected for forecasting the required outputs. Eight input parameters: density, cement, sand, sand-to-cement ratio, water-to-cement ratio, sand size, foam, and agent, were used to predict the CS of FC. Further, SHapley Additive exPlanations (SHAP) analysis was performed for feature interaction. The results indicate that the RF model had higher accuracy in forecasting the CS of FC. The assessment metrics, including R 2 values of 0.952 for 7 days and 0.958 for 28 days strength, and the mean absolute errors (MAE) of 1.47 MPa and 1.71 MPa, respectively, verified the results. A user-friendly graphical user interface (GUI) was developed to predict the CS of foam concrete using a machine learning model based on key input parameters. The impact of the study's findings is of great importance in reducing project costs by avoiding costly laboratory work and speeding up achieving desired outcomes.