An efficient numerical method for solving time-fractional differential equations using least squares support vector regression

Masoumzadeh, Reza; Biranvand, Nader

doi:10.22124/jmm.2026.32521.2951

An efficient numerical method for solving time-fractional differential equations using least squares support vector regression

Articles in Press

Document Type : Research Article

Authors

Reza Masoumzadeh ¹
Nader Biranvand ²

¹ Department of Applied Mathematics, Faculty of Mathematics and Computer Sciences, Amirkabir University of Technology (Tehran Polytechnic), No. 424, Hafez Ave., Tehran 15914, Iran

² Department of Mathematics, Faculty of Basic Sciences, Imam Ali University, Tehran, Iran

10.22124/jmm.2026.32521.2951

Abstract

In this paper, a numerical method is developed for solving time-fractional differential equations. By applying the weighted and shifted Grünwald-Letnikov approximation, the time-fractional diffusion equation is transformed into a semi-discrete scheme. Subsequently, least squares support vector regression (LS-SVR) and spectral methods are applied to solve the resulting semi-discrete scheme. The collocation LS-SVR approach is proposed for training the network. The resulting optimization problem is subsequently simplified to a linear system.

Keywords