GENETIC PROGRAMMING ALGORITHM AND ITS CLASSICALCOMPARISON BY NUMERICAL METHODS

Abstract

This article presents a theoretical and comparative analysis of Genetic Programming (GP) algorithm alongside classical numerical methods. The relevance of this research lies in the growing importance of automated discovery of mathematical relationships from empirical data in modern computational sciences. Genetic Programming, rooted in the evolutionary computation paradigm, evolves populations of mathematical expressions through crossover, mutation, and selection operators to find the best-fitting formula for given data. In this study, a GP algorithm was implemented from scratch in the Python programming language, and a web-based platform was developed for comparative analysis with the least squares method (LSM), Newton-Raphson, bisection, and secant methods. Experimental results demonstrate that GP significantly outperforms classical polynomial approximation methods in identifying complex nonlinear functions, although LSM retains advantages in speed and simplicity for straightforward polynomial relationships. The findings provide a scientific foundation for incorporating GP algorithms as supplementary tools in numerical methods curricula.