Studies on Instance Based Learning Models for Liquefaction Potential Assessment
Sudhirkumar Vinayakbhai Barai
Former Graduate Student
Liquefaction is a process by which sediments below the water table temporarily lose strength and behave as a viscous liquid rather than a solid. This causes a great damage to the structures constructed on those sediments. To assess the liquefaction potential, many models have been proposed. Earlier regression models were used to predict the liquefaction of sand deposit but those models were not consistent. Recently neural networks were used for such problems. In this paper an existing Instance Based Learning (IBL) is explored to predict the liquefaction potential. This is a machine learning approach, which creates an instance base of previous case records and predicts the result on the basis of its nearest(s) instance from the base. The IBL model is tested on Cone Penetration Test (CPT) dataset. The IBL performance showed improvement in case of CPT dataset over existing neural networks model.
Keywords: Cone Penetration Test, Instance Based Learning, Liquefaction, Machine Learning, Neural Networks
Get the entire paper