Prediction of swelling pressure of expansive soil using machine learning methods

Expansive soils present significant challenges in construction engineering due to their ability to swell, leading to structural damage. Accurate prediction of swelling pressure is essential for safe construction designs. This study compares the performance of four machine learning models, that is, Multiple Linear Regression (MLR), Decision Tree Regression (DTR), Support Vector Regression (SVR), and Artificial Neural Networks (ANN) in predicting swelling pressure based on soil properties such as KCl concentration, dry density, and moisture content. The ANN model outperformed the others, achieving the highest prediction accuracy with an R² of 0.92 and a mean squared error (MSE) of 0.0008 on the test data. SVR also performed well, with an R² of 0.91 and an MSE of 0.02. In contrast, the MLR model had an R² of 0.82 and an MSE of 0.002, while DTR had the lowest generalization capability with an R² of 0.54 and an MSE of 0.02. Although black box models like ANN and SVR are less interpretable, they significantly outperform traditional models like MLR and DTR in terms of prediction accuracy. This study highlights the potential of machine learning techniques in construction engineering and suggests that black box models, particularly ANN, can be highly effective in predicting swelling pressure with greater precision.