Data-Driven Modeling of Strength Properties in Waste-Stabilized Soils: A Review of Artificial Intelligence Approaches

The inclusion of industrial and agricultural by-products into soil stabilization offers a sustainable alternative to traditional binders like cement, which have a large environmental impact. Nonetheless, the strength behavior of soils treated with waste is not easily predictable due to complex nonlinear relations among soil constituent materials and the stabilizing agent(s). This presents a systematic review of the convergence of two research areas: (1) use of different waste products to improve geotechnical performance, and (2) application of artificial intelligence (AI) techniques to model soil behavior. This review synthesizes evidence on the use of materials such as fly ash, rice husk ash, refined granulated blast-furnace slag, and alum sludge, assessing the influence on important mechanical parameters, generally unconfined compressive strength (UCS) and California Bearing Ratio (CBR). Advanced AI and machine learning capabilities, such as Artificial Neural Networks (ANN), Genetic Programming (GP), and Extreme Gradient Boosting (XGBoost). Findings reveal that AI-based frameworks yield high accuracy (R2 > 0.95) and also frequently outperform empirical methods, while offering the strength of data integration. With some limitations concerning data availability, interpretability, and field applicability, AI-driven methods present considerable promise in the improved performance of soil-waste unit soils and more circular economies in geotechnical engineering.