Model for prediction of saltwater intrusion based on partial mutual information and fixed size least squares support vector machine

A model based on partial mutual information and fixed size least squares support vector machine is proposed, which used to deal with multivariable and large scale regression problem for prediction of saltwater intrusion,. The partial mutual information is used to select the inputs and its lag orders. A sparse approximation of input samples is done with an active selection of support vectors based on quadratic Renyi entropy criteria and Nyström approximation. The efficacy of the proposed approach is demonstrated on the East River.