The Secretary Problem is a well-known problem in optimal stopping theory that aims to determine the best strategy for selecting the best option from a pool of candidates. In this study, we focused on three sub-cases of the standard version of the Secretary Problem, which vary based on the availability of candidates' scores: no-information, full-information, and partial-information. First, we validated the existing theoretical solutions by conducting Monte Carlo simulations on generated data for the no-information and full-information cases which constitute the lower (0.37) and the upper (0.58) limits of the win rate, respectively. Then, we developed a Bayesian solution for the partial-information case which continuously learns the distribution of the scores from the interviewed candidates. The simulation results for the proposed solution showed that it is possible to reach the upper win rate limit without knowing any prior information about the population. We also tested the solution on a real dataset (Dutch exam scores) and achieved 0.53 win rate. We expect the proposed algorithm will provide a better solution than the commonly used no-information solution (37% rule) for the real-life applications of the Secretary Problem.