Researches On Deep Learning Based Recommendation Algorithms In E-learning System

With the progress of "Internet+Education" program. E-learning system become the first choice to obtain knowledge for many people. But there are some disadvantages in e-learning system like lack of personalized service. Using personalized recommendation technology to build an intelligence e-learning system become one of researching hotspots currently.The deep learning algorithms are introduced to build high-quality recommendation algorithms for learning resources. The new algorithms improve recommendation results effectively. The personalized recommendation technology could excite students’ learning interest and motivation.The work of this paper includes mainly two aspects as followed:(1) This paper makes a detailed analysis and summary on the traditional recommendation algorithms. In order to improve the robustness of recommendation algorithms on sparse data, an expanded autoencoder recommendation framework(Supervised Neural Recommendation, SNR) is proposed. The stacked autoencoders model is employed to extract the feature of input then reconstitution the input to do the recommendation. Then the side information of items and users is blended in the framework and the Huber function based regularization is used to improve the recommendation performance. Experimental results in terms of quantitative assessment show significant improvements over conventional methods.(2) In order to overcome the "cold start" problem in traditional recommendation algorithms, a content-based recommendation algorithm based on convolutional neural network (CNN) is proposed. The CNN can be used to predict the latent factors from the text information of the multimedia learning resources. To train the CNN, its input and output should first be solved. For its input, the language model is used. For its output, we propose the latent factor model, which is regularized by L1-norm.Furthermore, the split Bregman iteration method is introduced to solve the model.Experimental results in terms of quantitative assessment show significant improvements over conventional methods. And the split Bregman iteration method which is introduced to solve the model can greatly improve the training efficiency. The major novelty of the proposed recommendation algorithm is that the text information is used directly to make the content-based recommendation without tagging.