Aiming at the personalized movie recommendation problem, a recommendation algorithm in-tegrating manifold learning and ensemble learning is studied. In this work, manifold learning is used to reduce the dimension of data so that both time and space complexities of the model are mitigated. Meanwhile, gradient boosting decision tree (GBDT) is used to train the target user profile prediction model. Based on the recommendation results, Bayesian optimization algorithm is applied to optimize the recommendation model, which can effectively improve the prediction accuracy. The experimental results show that the proposed algorithm can improve the accuracy of movie recommendation.