Akademik Veri Yönetim Sistemi
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING, ss.1-18, 2026 (SCI-Expanded, Scopus)
This study focuses on the prediction of the plasticity index of fine-grained soils using machine learning methods, such as support vector regressor, decision trees, random forest, gradient boosting regression, extreme gradient boosting, and multilayer perceptron, based on liquid limit and plastic limit parameters. A comprehensive dataset of 341 data points with liquid limit values between 19.00 and 140.60 was generated from the literature. This dataset was then split into an 80% training set and 20% testing set. Notably, support vector regressor and multi-layer perceptron models demonstrated superior prediction accuracy and generalization capability, with support vector regressor reaching a determination coefficient of 0.999 and the lowest error rates. The Shapley additive explanation analysis indicated that the liquid limit played a more crucial role in the predictions than the plastic limit, highlighting its importance in estimating the plasticity index. Polynomial regression equations were developed for a simplified estimation of the plasticity index from the liquid limit for different soil types. The findings showed improved accuracy when the low- plasticity clay and high-plasticity clay soil data were evaluated together. Compared to previous studies, support vector regressor model outperforms existing models in terms of accuracy and error minimization. The study concludes that machine learning models, especially support vector regressor and multilayer perceptron, present a fast, accurate, and low-cost alternative for predicting the plasticity properties of clayey soils, and recommends future studies to expand datasets and include additional soil parameters for enhanced model performance.