To realize personalized and high-quality resource recommendation of online learning platform, this paper proposes a collaborative filtering algorithm based on RFM-K-means clustering. Firstly, this algorithm calculates user-product analysis matrix by RFM model and obtains weight of each index in RFM by entropy weight method. Then, K-means clustering is utilized to extract user features. In addition, based on user-based collaborative filtering method (User-based CF), resources are recommended. Finally, by constructing user-learning scoring matrix and calculating user similarity to generate nearest neighbor set, personalized resource recommendations can be achieved. The results show that based on the data source of an online learning platform, accuracy of the test questions with a prediction score of not less than 0.6 recommended by the proposed recommendation model to target users is 90%, showing high prediction accuracy. Thus, collaborative recommendation method constructed in this study can realize the personalized resource recommendation of online learning platform, which has certain application value.