Peer assessment is an effective evaluation method for online courses, which can solve the problem of large-scale online learning evaluation and improve the evaluation efficiency. Aiming at the problem of matching in peer assessment, a reviewer matching recommendation strategy based on training is proposed. According to the analysis of training data, it mines the evaluation characteristics of learners and generates the peer assessment recommendation list. The practice result shows that this strategy can ensure the effectiveness and reliability of peer assessment and promote learners' learning introspection