Answers are grounded in this paper and cite their sources.
In domains such as the construction industry, predicting the compressive strength of laterized blocks is crucial for building trust in their results. Poor material properties can result in reduced strength, durability, and structural integrity of buildings and can have even life-threatening consequences for building occupants. Machine learning (ML) algorithms can provide predictive abilities for compressive strength of laterized blocks, thereby improving confidence in their use in buildings. This research aims to predict the compressive strength of laterite blocks stabilized with metakaolin-based geopolymer (MKG) and sugarcane molasses (SM) using gradient boosting regression (GBR), multi-layer perceptron (MLP), k-means clustering, and AdaBoost models. The correlation matrix results show that stabilized compressed earth block (CEB)'s age has the strongest correlation with compressive strength (R = 0.83) based on its highly meaningful relationship (p < 0.001). The research establishes the credibility of the ML algorithms during the training phase through R-squared values higher than 0.93 for GBR, MLP, k-means clustering, and AdaBoost models. However, the R-squared values during the testing phase are less than 0.68 for all the models. The Taylor diagram analysis indicates that the GBR model is the best-performing model, while the k-means clustering model emerges as the lowest-performing model during the training phase. Based on mean SHapley Additive exPlanations (SHAP) analysis values, age emerges as the most significant variable in influencing the compressive strength of laterite blocks stabilized with MKG and SM mixtures. Moreover, water, MKG, and SM have also emerged as influential variables in the order of decreasing influence. This research therefore lays a foundation and contributes to the widespread applications of laterite blocks stabilized with MKG and SM mixtures in sustainable construction.
No takes yet. Share an insight, caveat, or question.
Sinkhonde et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: