Energy-Based Least Squares Projection Twin SVM

Abstract

In this paper, we propose energy-based least squares projection twin support vector machines (ELSPTSVM) for classification problems. The proposed ELSPTSVM models seek projection directions for each class in such a manner that the data of each class are well separated from the data of other class in its respective subspace. To further enhance the performance, the proposed ELSPTSVM model uses a recursive procedure to generate multiple projection axis for each class. The proposed ELSPTSVM model uses energy parameters and solves a system of linear equations to boost the classification performance. The experimental results show the efficiency of the proposed ELSPTSVM model on benchmark UCI datasets with binary class and some non-UCI datasets of fecundity estimation for fisheries data of binary class. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.