Research On Deep Matrix Factorization And Intelligent Recommendation Methods For Learning Resource Adaptation

In the process of fighting against novel coronavirus pneumonia,online education has been developing rapidly.While playing an important role in knowledge acquisition,it has gradually become a new way for learners to gain new knowledge.However,the development of big data in education has caused problems of information redundancy and uneven quality of educational resources.When choosing educational resources,learners not only have to face the problem of choosing the right educational resources for their needs from numerous ones,but also have to distinguish and choose the high-quality ones among the proper ones.Undoubtedly,this is quite difficult for learners.Therefore,the urgent problem of choosing high-quality educational resources suitable for learners from the vast ones needs to be solved.Personalized recommendation method is an efficient and quick solution with long-term development prospect,which can effectively extract the preference characteristics of learners and select appropriate high-quality educational resources for them.In this thesis,deep learning technology is used to model learners’ feedback information and build an efficient resource recommendation algorithm,which can effectively improve the accuracy of the recommendation results.It enables learners to get a more personalized learning experience in the process of online learning.This thesis mainly proposes two models to solve the learner’s choice predicament in the online environment.The main work of this thesis is as follows:(1)a novel confidence-aware recommender model via review representation learning and historical rating behavior is presented in this thesis.The interaction latent factor of user and item in the framework is constructed by exploiting review.information interactivity.Then,the confidence matrix,which measures the relationship between the rating outliers and misleading reviews,is employed to further improve the model accuracy and reduce the impact of misleading reviews on the model.Furthermore,the loss function is constructed by maximum a posteriori estimation theory.Finally,the mini-batch gradient descent algorithm is introduced to optimize the loss function.Experimental results demonstrate that the proposed model achieves state-of-the-art performance on several public datasets.(2)an efficient deep matrix factorization is proposed with L0-regularized review feature learning for recommendation systems(EDMF).Two characteristics in user’s review are revealed:Firstly,the interactivity is exploited in a review between the user and the item,which can also be considered as the user’s scoring behaviors on the latter;Secondly,the review is only a partial description of the user’s preferences for the item,which is revealed as the sparsity property.Regarding the first characteristic,EDMF extracts the interactive features of single review by convolutional neural networks with word attention mechanism.Then,considering that the review information is a sparse feature,which is the second characteristic,the L0 norm is employed to constrain the review.Furthermore,the loss function is constructed by maximum a posteriori estimation theory.Finally,to optimize the loss function,the alternative minimization algorithm is introduced.Experimental results on several datasets demonstrate the proposed methods outperform the state-of-the-art methods in term of the effectiveness and efficiency,which show good education application prospects.