Research On Knowledge Recommender Method Based On Transfer Learning

E-learning have many advantages over traditional learning since they often have a clear mandate and the learning process of learners with the ability to grow with knowledge. But how to match individual learning needs among vast knowledge resources is still a significant challenge. Recommendation algorithm is an effective way to solve the problem, but since the particularity of internet knowledge learning environments the existing recommendation algorithm is unable to meet the stepwise learning needs of learners and cannot reach the required mining interests and learning potential of learners.The main contents of this paper are: 1) Research on behavioral modeling of learner and knowledge resource modeling in internet knowledge learning environment, in this paper we focus on considering learners preferences and their level of capacity in each areas, the domain that knowledge resource belongs and its difficulty level and other characteristics; 2) We combine the learner model and the knowledge resource model with the existing recommendation algorithm to get the recommendation algorithm that applicable to the internet knowledge learning environment; 3) Combining the "mobility" in learning behavior of human being, we try to make use of transfer learning method to design knowledge recommendation algorithm, so that the accuracy of the knowledge recommendation algorithm can be increased, and the interest and potential of the learner can be dug out.In this thesis we firstly put forward an algorithm named Item-Based Transfer Learning. And then we did the verified experiment with the public data; the experiment result shows the method is effective, and when the auxiliary data set is dense and the target set is sparse, the effect is obvious. We also proposed a knowledge recommendation algorithm framework based on transfer learning, and we applied the framework in online judge system, and the validity and applicability of the framework is verified by experiments. The research will help promote the development of internet knowledge learning and provide an effective solution for the future construction of a more intelligent, personalized internet knowledge learning environments.