Employing hybrid meta-heuristic algorithm for nonlinear regression model estimation

Abstract

Modeling of stochastic phenomenon is widely done using nonlinear regression models and the estimation ofparameters problem plays a central role in the inference in nonlinear regression models Nonlinear regression models areapplicable in solving optimization problems.As the traditional techniques would fail to produce the satisfactory results inparameters estimation problem.The methods that have to be taken into consideration are Newton-Raphson Method (NRM),Gauss-Newton Method (GNM), The Levenberg-Marquardt Method(LMM), Quasi-Newton methods (QNM), DavidonFletcher Powell Method (DFP), Broyden Fletcher Goldfarb Shanno Method(BFGSM) and Nonlinear Conjugate GradientMethod (NCG), In this paper we propose A new hybrid algorithm Harris hawks optimization (HHO)that can enjoy thestrengths of both the descent algorithms and the natural optimisations algorithms. Via computer simulation, it has beendemonstrated how the (HHO-NRM), (HHO-GNM), (HHO-LMM), (HHO- QNM), (HHO-DFPM), (HHO -BFGSM) and(HHO-NCGM) hybrid algorithm- based formula that is powered by the conventional algorithms greatly outperform theconventional algorithms by squared error rates. This was done by use of nonlinear functions. The computational findingsindicate that the summed up HHO is favorable over to the numerical techniques (NRM), (GNM), (LMM), (QNM), (BFGSM)and (NCGM) in majority of the problem formulations used in this work.