Research On Personalized Learning Recommendation Based On Incremental Collaborative Filtering

In recent years, with the rapid development of Internet and network learning community, information technology is changing people's ways in learning at an amazing speed. At the same time,that the problems of learning trek,knowledge overloaded gradually revealed. Personalized learning recommendation has become one of the important means for people to quickly learn and grasp the knowledge. And it reflects the learner centered personalized autonomous learning mode and effectively improved the problem of learning and knowledge overload trek in a certain extent. Through the analysis of the learning characteristics and learning behavior record information, personalized learning recommendation system can recommend service to the learners that they may be interest. And the personalized learning recommendation has become a research hot spot of social learning network and educational data mining.At present, collaborative filtering is one of the most widely used and mature algorithms in the personalized recommendation system. However, the recommendation algorithm based on collaborative filtering, which needs further research and solution, still faces some bottleneck problems, such as sparsity, scalability, cold start and so on. Aiming at the problem of sparse data, scalability and accuracy, this paper proposes a personalized recommendation algorithm which based on incremental collaborative filtering. The main works of this paper are as follows:Firstly, this paper proposed a multi agent collaborative scoring model driven by double layer behavior. Through the learning social network relationship, dynamic agent perception and the ability to interact, the model will predict the prediction score which can reduce the deviation by combining the interest degree of the individual behavior, the trust degree and the influence of group behavior measurement.Secondly, according to the sparsity, scalability and accuracy of traditional collaborative filtering algorithm, a new method based on incremental collaborative filtering is proposed. By using multi agent collaborative scoring model which is double behavior driven, the interference can be reduced by sparse data, and the reliability of scores is also improved. In order to improve the efficiency and accuracy of the recommendation algorithm, that the dynamic K-means clustering to dynamically partition the sample space and modify collaborative filtering and incremental update mechanism is used to extract the recent neighbor problem.Finally, experiment on CSDN data set is done to compare the results. It is proved that multi agent collaborative scoring model driven by double layer behavior can improve the accuracy of the collaborative algorithm. Experimental results show that the algorithm proposed in this paper has higher accuracy, better performance and scalability.