<p>Initial Orbit determination (IOD) is a critical step in cataloging Near-Earth Asteroids (NEAs). Traditional IOD methods, such as the Laplace, Gauss, and double-r iteration methods, typically generate multiple solutions. However, there has been limited research on selecting the optimal solution from these multiple initial orbit solutions. This paper proposes a multi-criteria integration method for selection of optimal initial orbit solution. In addition to using observational angular residuals, we introduce three new criteria: the variance of absolute magnitude, angular velocity deviation, and the joint probability density of orbit elements. An optimization algorithm based on a continuous orbit-distance loss is designed to determine the weight coefficients for each criterion, while the original three-level score is retained only as an a posteriori accuracy classification. Both training and testing datasets were generated using real NEA data from the Minor Planet Center. Numerical simulations demonstrate that the proposed method effectively identifies the optimal solution from a set of multiple initial-orbit solutions, achieving a correct rate of 96.52% on the training set and 96.67% on the test set. Additional short-arc experiments further show that the method remains effective in operationally relevant single-night and two-night scenarios.</p>