Abstract
This study investigates using machine learning techniques to predict the compressive strength of cement-stabilized earth blocks (CSEBs). CSEBs are a promising sustainable construction material, but their compressive strength depends on various soil characteristics. Accurately predicting this strength is crucial for design and construction purposes. The research analyzes the influence of several soil properties, including particle size distribution, Atterberg limits, and compaction test results, on the compressive strength of CSEBs. For this purpose experimental program was conducted using nine different soils and three different cement contents to prepare the CSEBs. Additionally, it explores the efficacy of an Artificial Neural Network (ANN) model in predicting this strength based on these soil characteristics. The findings reveal that cement content significantly impacts compressive strength, followed by other factors like the coefficient of curvature, sand content, and liquid limit. Utilizing SHAP (SHapley Additive exPlanations) analysis allows for interpreting the model and identifying the key features influencing its predictions. Focusing on a reduced set of crucial features identified through SHAP analysis can maintain acceptable prediction accuracy while reducing data acquisition efforts. This research signifies the potential of machine learning, particularly ANN models, for accurately predicting the compressive strength of CSEBs based on their soil properties. This advancement can contribute to the efficient and sustainable development of constructions utilizing CSEBs.