To address the current research limitations in the probabilistic stability analysis of unsaturated embankments，this study proposes a novel probabilistic analysis approach for embankments stability，combining a selflearning strategy with the Gaussian Process Regression（GPR）model. By considering the variabilities of soil parameters within embankment，the GPR model was used to establish the relationship between the random input variables and the minimum safety factor（FSmin）of embankment. An active learning strategy was employed to determine the optimal number of training samples，thereby significantly enhancing the computational efficiency. Validation through an embankment case demonstrates that the proposed method achieves a computational efficiency approximately 125 times higher than traditional Monte Carlo Simulations（MCS）method. The proposed method provides an efficient and accurate tool for the probabilistic analysis of unsaturated embankments stability，accounting for uncertainty factors.