Predicting Fredlund–Xing soil-water characteristic curve parameters from grain-size and Atterberg-limit data: algorithm choice, soil grouping and input information
Year: 2026
Type: article
Abstract: Code, training data and trained models for the paper “A Controlled Benchmark of Machine-Learning Models for Soil–Water Characteristic Curve Prediction: Algorithm сhoice, Soil Grouping, and Predictor Importance” (Yerzhanova et al., Nazarbayev University). The study predicts the Fredlund and Xing (1994) SWCC parameters a, n and m from basic soil properties using ANN, GBR and RF models. Each algorithm is trained as fine-grained, coarse-grained and global models, validated on an independent set of 90 soils, and analysed for predictor importance. The repository includes the training datasets (derived from the USDA NCSS database), training and validation code, saved train/test splits, trained models and a script that predicts the full SWCC for a new soil.
Language: English
Topic: Soil and Unsaturated Flow
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7222062299