Machine Learning-Based Prediction of Soil Compaction Parameters

Abstract This study aims to predict the soil compaction parameters maximum dry density (MDD) and optimum moisture content (OMC) using machine learning models. As a result of this aim, fast and much less costly advanced machine learning models can be used as an alternative to traditional methods such as proctor test, which are time-consuming and costly in geotechnical engineering. The study conducts a comparative analysis of various machine learning models such as decision tree, random forest, gradient boosting and group method of data handling (GMDH) to determine the most accurate and efficient model. As a result of this comparative analysis, the random forest model was the most successful model. According to the 5-fold cross-validation results of the random forest model, the average training coefficient of determination (R²) value is 0.95 and the average testing R² value is 0.71. Surely, if we had a dataset with more data, higher performance results could be achieved with existing machine learning models. The results obtained demonstrate that machine learning models increase project efficiency and reduce costs by reducing reliance on laboratory testing, and offer an approach with the potential to improve soil compaction applications in geotechnical engineering.