NEURAL NETWORK-DRIVEN ADAPTIVE ROOT-FINDING ALGORITHM: LEARNING TO SOLVE NONLINEAR EQUATIONS MORE EFFICIENTLY

Root-finding methods such as Bisection, Newton-Raphson, and Secant are classical algorithms for solving nonlinear equations. This study proposes an adaptive classification-based approach using a neural network to predict the best method based on initial iteration behavior. The model achieves up to 100% classification accuracy. Results are validated with training/test loss trends, classification reports, and confusion matrices, and benchmarked against classical numerical analyse.