Machine Learning Methods for Predicting Soil Compression Index

The compression index is an important consideration when figuring out how fine-grained soil settles. The compression index is determined from the oedometer consolidation test which is tedious and time-consuming. As a result, numerous correlations between the compression index and the index properties were developed. As soil is a very unpredictable substance, those correlations do not hold for all types of soil. This opens the door for the development of machine learning methods to forecast compression index. In this study, the compression index of soil is predicted using a decision tree, random forest, and multiple linear regression. Index properties, like liquid limit, natural moisture content, initial void ratio, and plasticity index are used as input variables in the machine learning models that are created to forecast the output variable compression index. The dataset used contains 359 data from diverse soil types and was gathered from several published articles (CH soil—62, CI soil—186, and CL soil—111). Since the machine learning models are trained using the training dataset before being evaluated using the testing dataset, the data has been divided into a training dataset and a testing dataset. In this paper, the impact of data splitting is also examined because it affects model performance. Mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R²) are used to assess the performance of the models. The results show that when training, decision trees perform well, whereas the testing dataset favors multiple linear regression for prediction. The data partitioning that results in the optimum performance for each model is different.