Despite the fact that online course recommendation system plays a key role in personalized learning path design, traditional recommendation algorithms, which often adopt fixed weight settings, fail to flexibly adapt to the changes of users’ interests. In order to address this issue, firstly, a multi-task recommendation algorithm WAA-TA has been proposed with weighted auxiliary task perception integrated. The algorithm dynamically adjusts task weights and uses task sets to represent users’ learning needs at different stages of their lifecycle, thus improving the accuracy and personalization of the recommendation system. Secondly, based on comparative experiments conducted with six baseline algorithms on the Edx and MOOCCubeX datasets, the experimental results show that this algorithm is characterized by an excellent performance in various evaluation indicators, especially in improving user satisfaction and recommendation accuracy.