The combination of explicit and implicit feedback can effectively improve recommendation performance. However, the existing recommendation systems have some disadvantages in integrating explicit feedback and implicit feedback, i.e., the ability of implicit feedback to reflect hidden preferences from missing values is ignored or the ability of explicit feedback to reflect users' preferences is not fully utilized. To address this issue, this study proposes an explicit and implicit feedback based collaborative filtering algorithm. The algorithm is divided into two stages, where the first stage deals with implicit feedback data by weighted low rank approximation to train implicit user/item vectors, and the second stage introduces a baseline estimate and uses the implicit user/item vectors as supplementaries to the explicit user/item vectors. Through the combination of explicit and implicit user/item vectors, the predictions of users' preferences for items can be obtained by training. The proposed algorithm is compared with several typical algorithms on standard datasets, and the results confirm its feasibility and effectiveness.