A Knowledge Recommendation Algorithm Based on Time Migration<bold> </bold>

With the rapid development of the Internet, the transmission and acquisition of information is becoming more and more convenient, and the problem of information overload is becoming more and more serious. The birth of the recommendation system effectively alleviates the problem of information overload, which can recommend potential interests to the users according to the user characteristics and project characteristics of the system. So far, the recommendation system has been widely used in business, but it is still immature in the field of online learning. Time transfer as one of the important influencing factors of recommendation system, its influence mode still needs further research and experimental investigation. To solve this problem, we analyze the users' behaviors, and propose time migration model based on different behaviors of users: short-term behavior model and longterm behavior model. With the data of online learning system, modeling the relationship between learners, problems and learners-problems, we propose a knowledge recommendation algorithm based on time migration (KRBTM). Our approach involves four steps: (1) creating the time migration model, (2) adjusting the model of learners and learners-questions with time migration model, (3) computing ratings similarity of questions based on time migration model, (4) applying KRBTM to recommend the top-N questions to learners. In this paper, we test the time migration model on Movielens with a good result. We also apply KRBTM to the real datasets of TJU ACM-ICPC Online Judge from Tianjin University, and the experiments show that KRBTM has a higher recall and F value than traditional algorithms.<bold> </bold>