Proceedings of the
8th International Symposium on Geotechnical Safety and Risk (ISGSR)
14 – 16 December 2022, Newcastle, Australia
Editors: Jinsong Huang, D.V. Griffiths, Shui-Hua Jiang, Anna Giacomini, Richard Kelly
doi:10.3850/978-981-18-5182-7_08-003-cd
Prediction of Parallel Desiccation Cracks of Clays Using a Classification and Regression Tree (CART) Technique
1School of Engineering, Deakin University, 3086 VIC, Australia.
2School of Engineering, IT and Physical Sciences, Federation University Australia, 3842 VIC, Australia.
3WA School of Mines: Minerals, Energy and Chemical Engineering, Curtin University, Kalgoorlie, WA, 6430, Australia.
ABSTRACT
One of the most effective and common phenomena, especially in arid and semi-arid regions, is cracking caused by soil drying. Cracking changes soil properties such as permeability. As a result, recognizing the amount and nature of cracks in clay and predicting them can greatly contribute to infrastructure design. This paper seeks to predict the number of parallel clay cracks using a classification and regression tree (CART) technique under several variables including initial water content, soil thickness and sample width and length as input parameters. A database consisting of 31 datasets was used in the study. To evaluate the accuracy of the model, two statistical indices, namely the coefficient of determination (R2), root mean square error (RMSE) have been used. The results are compared with the Multiple linear regression (MLR) method. Results show that for testing database, R2 and RMSE based on the classification and regression tree (CART) model are 0.989 and 1.285, respectively, while the R2 and RMSE in the multiple linear regression method were equal to 0.777 and 41.696, respectively. As a result, the classification and regression tree (CART) has performed well in predicting the number of desiccation cracks.
Keywords: Parallel desiccation cracking, Clay, Artificial Intelligence, Classification and Regression Tree.
