This study presents a machine learning approach using least squares support vector regression (LS-SVR) for solving fractional pantograph diferential equations involving the left-sided Bessel fractional (LSBF) derivative. First, a new orthonormal basis, called the fractional-order airfoil functions (FAFs) is constructed. Then, the unknown function and its derivatives are approximated via FAFs. The collocation LS-SVR method is applied to train the network with the FAF kernel. The method’s formulation results in an optimization problem that an equivalent system of algebraic equations is derived. Finally, the error analysis of method is presented and numerical tests are provided to show the efciency of the suggested technique.
