Research On Learning Resource Recommendation Algorithm Based On Collaborative Filterin

With the development of educational informatization,online learning is gradually favored by learners due to its unique advantages.However,when learners use online platforms,it is difficult to choose from a large number of learning resources.It can be seen that there is an urgent need to recommend learning resources.Personalized recommendations can alleviate the above problems,so that learners can obtain appropriate learning resources according to their actual needs.The current focus of domestic and foreign scholars is how to choose the appropriate recommendation technology.Adapting recommendation results to user preferences,so as to better improve the user’s personalization level.This thesis aims to deeply discuss the learning resource recommendation algorithm based on Collaborative Filtering(CF).The main research content is divided into the following three aspects:(1)A collaborative filtering recommendation algorithm based on multi-similarity fusion is proposed.First of all,considering global and local similarities,the similarity measurement method is improved accordingly,GL(Global and Local)is proposed,which overcomes the limitations of the common-evaluation items.Considering the influence of non-common-evaluation items involving valuable potential information and improving the practicability of the calculation results.Then,based on the K-Nearest Neighbor(KNN)algorithm,a recommendation algorithm KNN-GL optimized is proposed.Secondly,based on the learner’s interest model,a learning resource recommendation method integrating the Bisecting K-means and KNN-GL is proposed.The new learning resource recommendation method achieves accurate recommendation of learning resources for learners.Finally,the experimental results show that the algorithm in this thesis is obviously helpful to improve the recommendation effect,and alleviates the cold start and data sparsity problems of traditional recommendation algorithms to a certain extent.(2)A collaborative filtering recommendation algorithm based on an online learning style model is proposed.First,considering the concept of learning style,focusing on group-based collaborative learning,an online learning style model is constructed.Then,the K-means algorithm is introduced to filter homogeneous samples through clustering while reducing the computational cost,and integrates the Apriori algorithm and the KNN algorithm to extract useful information from the log files of learners in each cluster and calculate the association rules between learning resources.It is used to further calculate the similarity between learning resources between clusters.At the same time,an adaptive recommendation method CFROLS is proposed.The new learning resource recommendation method provides learners with more personalized recommendation services.Finally,through the design experiment of the OULAD dataset,the results show that the performance of the algorithm is relatively stable due to the limitation of the length of the recommendation list.(3)A collaborative filtering recommendation algorithm incorporating generalized Sequential Patterns(SPM)is proposed.First,a new scoring function is defined based on mining web log files to weight learning resources as objects,taking into account learners’ explicit feedback and implicit preferences.Then,the KNN algorithm is used to obtain the most suitable recommendation list from the resource library,and the Sequential PAttern Discovery using Equivalence classes(SPADE)algorithm is introduced to construct learning scenarios to achieve personalized recommendation.Therefore,a learning resource recommendation method that integrates collaborative filtering and sequential pattern mining is proposed,which is more suitable for learners’ learning needs.Finally,an experiment is designed to evaluate prediction accuracy by using an external dataset in the ELearning environment,which shows that using this method can improve the prediction performance,and at the same time proves the scientificity and effectiveness of this method.