Research On E-learning Resource Recommendation Method Based On Deep Learning

With the explosive growth of network learning resources and the digitization of a large number of traditional learning resources,people’s learning style has gradually changed into learning through the network.However,there are many online learning resources on the Internet.A learner often needs a recommendation system to recommend suitable online learning resources for him.However,the current online learning resource recommendation methods can not fully extract the relationship between learners and online learning resources,which makes the recommendation results unsatisfactory.This paper aims to improve the effect of recommendation and solve the problem of cold start in online learning resource recommendation:1.An online learning resource recommendation method based on G-LSTM.G-LSTM model combines Generalized Matrix Factorization and Long Short-Term memory network,which makes it possible to deal with the temporal data in online learning resource data and effectively capture the relationship between learners and online learning resources.This method mainly includes three steps.Firstly,the latent factor matrix representing all learners and the latent factor matrix representing all online learning resources are extracted by feature extraction of learners and online learning resources.Then,through the potential factor matrix extracted in the previous step,the eigenvectors of each learner and each online learning resource are calculated respectively,and they are input into GMF module and LSTM module of G-LSTM model in pairs.Finally,the output results of GMF module and LSTM module are combined by multi-layer perceptron,and the recommendation list is generated according to the results.A large number of experiments are carried out on the real data set provided by uniform college,and the recommendation ability of the method is evaluated with hit rate and normalized loss cumulative gain as evaluation indexes.2.An online learning resource recommendation method based on Generative Adversary Network.In this method,Deep Neural Network is used to construct generator model and discriminator model,and the generative relationship between learners and online learning resources is used to generate online learning resource data through generator model,so that Generative Adversary Network model can be trained to converge.After the training of the model,the result value generated by the discriminator is used to generate the recommendation list.The precision rate and recall rate are used as evaluation indexes to evaluate the recommendation ability of the method.