Comparison of Performance of Machine Learning Models for Predicting Compression Index Based on Clay Properties

As the construction of large structures increases primarily on soft ground along coasts, prevention damage to due subsidence should be prioritized. Ground has been investigated actively. Because amount settlement can calculated based compression index clay, latter must accurately. In this study, data pertaining natural water content, liquid limit, plasticity index, initial void ratio, and clay are acquired from Busan, Gwangyang, Mokpo construct a dataset for predicting index. Correlation analysis between factors is performed using dataset, prediction models developed machine learning algorithms, random forest, multiple linear regression, ridge, Lasso, SVM, XGBoost, LightGBM, DNN. Subsequently, results each model compared in terms RMSE R2. The show that correlated significantly. Among models, LightGBM demonstrates best performance.