Applications of machine learning algorithms on the compressive strength of laterite blocks made with metakaolin-based geopolymer and sugarcane molasses

• ML algorithms for predicting strength of laterite blocks made with metakaolin-based geopolymer (MKG) and sugarcane molasses (SM) are presented. • The research addresses the performance comparisons of the ANN, RF, DT and SVM models. • The Taylor diagram model and SHapley Additive exPlanations ( SHAP) analyses are also included. • Knowledge gaps and possible futher research considerations are identified. To refine the process of anticipating the structural integrity of laterite block components, the use of machine learning (ML) algorithms is required. This study initiates an exploration into forecasting the compressive strength of laterite blocks infused with metakaolin-based geopolymer (MKG) and sugarcane molasses (SM), utilizing machine learning techniques such as artificial neural networks (ANN), random forests (RF), decision trees (DT), and support vector machines (SVM). The models were developed using four input values, including the MKG, SM, laterite soil, and water, with compressive strength as the output. Results show that for all the models, the majority of the data points lie within the error lines range of −20 % and +20 %. Using the Taylor diagram model, the results demonstrate that the SVM (train) model achieves the highest performance in predicting the compressive strength of laterite blocks, with a correlation coefficient of 0.99 and the lowest root mean square error (RMSE) of 0.139. The correlation coefficient values (R) for training and testing algorithm models ranged between 0.65 and 0.99, implying that all models fairly predict the compressive strength of laterite blocks containing MKG and SM. The RF model emerges as an important model for generalization across training and testing phases, with R values of 0.9828 and 0.789, respectively. SHapley Additive exPlanations (SHAP) analysis assesses the model’s explainability behavior. According to a SHAP-based feature importance study, age (85.33 %) and water content (17.87 %) are critical components that may improve compressive strength compared to MKG (8.60 %) and SM (6.74 %), respectively. This study not only assists in comprehending the essential parameters necessary for making well-informed decisions but also opens exciting possibilities for the application of ML in fostering sustainable construction practices.