Predict the Maximum Dry Density of soil based on Individual and Hybrid Methods of Machine Learning

Document Type : Original Article


1 Department of Mechanical Engineering, School of Technology, GSFC University, Vadodara, Gujarat, India

2 Department of Civil and Water Engineering, University of Tabriz, Tabriz, Iran

3 Australian Maritime College, College of Sciences and Engineering, University of Tasmania, Launceston, 7248, Australia


This article introduces a novel technique to accurately forecast soil stabilization blends' maximum dry density (MDD). The Naive Bayes (NB) algorithm is employed to develop detailed and accurate models that use various natural soil characteristics, such as particle size distribution, plasticity, linear shrinkage, and stabilizing additives' type and amount, to relate to the MDD of stabilized soil. To ensure the model's accuracy, the study integrates two meta-heuristic algorithms: Artificial Rabbits Optimization (ARO) and Gradient-based Optimizer (GBO). The models undergo validation using MDD samples of various soil types acquired from previously published stabilization test results. The results reveal three distinct models: NBAR, NBGB, and an individual NB model. Among these, the NBAR model stands out with exceptional performance, boasting a high R2 value of 0.9903 and a remarkably low RMSE value of 34.563. These results demonstrate the precision and reliability of the NBAR model and signify its effectiveness in predicting soil stabilization outcomes. Overall, this approach offers a promising way to accurately predict the MDD of soil stabilization mixtures in various engineering applications. Integrating meta-heuristic algorithms into the analysis increases the accuracy of the models and provides more reliable predictions, which has significant implications for the construction industry, where soil stabilization is critical for building robust and long-lasting infrastructure.