The standard mean-variance portfolio selection model assumes that investors exactly know the seurity parameters，neglecting the effect of parameter uncertainty on portfolio selection．This paper investigates a continuous。time mean-variance portfolio selection problem under parameter uncertainty and Bayesian learning． The problem is solved by using a martingale approach，and the optimal investment strategy and the mean-variance efficient frontier are derived in closed form．Based on these results．we sive an empirical analysis with data from Chinese security markets．The analysis shows that parameter uncertainty has a great effect on the optimal investment strategy and the investment performance．