Linear kernel support vector machines for modeling pore-water pressure responses

• Main Authors: Yusof, K.W., Babangida, N.M., Mustafa, M.R., Isa, M.H.
• Format: Article
• Published: Taylor's University 2017
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85026916080&partnerID=40&md5=292625363e1b6d38bea120e0107fbbba; http://eprints.utp.edu.my/19783/

Pore-water pressure responses are vital in many aspects of slope management, design and monitoring. Its measurement however, is difficult, expensive and time consuming. Studies on its predictions are lacking. Support vector machines with linear kernel was used here to predict the responses of pore-water pressure to rainfall. Pore-water pressure response data was collected from slope instrumentation program. Support vector machine meta-parameter calibration and model development was carried out using grid search and k-fold cross validation. The mean square error for the model on scaled test data is 0.0015 and the coefficient of determination is 0.9321. Although pore-water pressure response to rainfall is a complex nonlinear process, the use of linear kernel support vector machine can be employed where high accuracy can be sacrificed for computational ease and time. © School of Engineering, Taylor’s University.