<p>Active learning Kriging has found extensive application in structural reliability analysis. One crucial aspect of this method is using learning strategies to identify optimal samples. This article aims to provide a two-stage active learning Kriging method based on portfolio allocation of learning strategies and importance sampling (IS). The first stage provides a Kriging model to estimate the failure probability, and the second stage gives a correction factor to quantify and modify the predictive error made by the inaccurate surrogate model. In Stage I, the proposed method leverages the portfolio allocation strategy to renew Kriging model. A novel reward function within portfolio allocation strategy is developed from the perspective of reducing the global approximation uncertainty. In Stage II, an instrumental IS density is first given to generate independent samples for estimating the correction factor. To reduce the computational cost, a new Kriging model is built in Stage II. The portfolio allocation strategy is further used for the renewal of this Kriging model. Two numerical examples and one engineering example are used to validate the performance of the portfolio allocation strategy and the proposed method.</p>