With the increase of online courses, online courses are facing the problem of information overload. This article focuses on solving the problem of the application of collaborative filtering algorithms in online courses. By analyzing the types of courses in the online courses and the behaviors of users, the online courses are classified to increase the high degree of cooperation of the courses. An item-based collaborative filtering algorithm is used, and user activity is penalized based on IUF in calculating similarity, so as to obtain a more accurate recommendation list. Experimental results show that the algorithm can improve the recommendation quality in the application of online courses.