Research On Personalized Resource Recommendation For Online Learning System

With the progress of information technology,online learning based on network has developed rapidly.The internet-based online learning system utilizes basic hardware and cloud technology to enable learners and educators to ignore the differentiation of time geography and learn more easily.It is not constrained by the characteristics of classroom learning,greatly satisfy the characteristics of contemporary education lifelong learning needs,also enriched the diversity of education.As the learning system of the online learning system is phased in,students can use the application platform to obtain end-to-end digital learning and generalization learning.The establishment of the online learning system enables the students to establish personal learning space and the learning activities,but the traditional remote teaching system and the auxiliary teaching system just provide online learning environment for learners to learn,does not take the personalized education mode into account,even can't take intellectualized teaching according to individual differences.When the traditional web recommendation algorithm was transplanted to the online learning system,the effect of personalized recommendation learning was poor due to the failure to consider the transfer of students' interest over time.In view of the shortcomings of traditional recommendation that interest changes over time,we introduce LDA time weighted collaborative filtering to compute user's interest probability distribution,and combines the time interest factor to calculate the similarity between users to complete the collaborative recommendation;In order to solve the cold start problem of cooperative algorithm and examine the relationship between knowledge,the knowledge unit closure is introduced to complete the final recommendation analysis with the former.Based on the analysis of various traditional recommendation algorithm,we combine interest model with the time dimension and knowledge unit closures,then put forward the suitable personalized hybrid recommendation algorithm for online learning system based on the LDA and association rules.The algorithm is able to obtain the students' learning objectives and interests by analyzing the students' learning path topics and obtain the distribution of each topic through the Gibbs sampling,and get the distribution of the subject corresponding learning courses and construct the minimum knowledge unit closure by the association rules algorithm.Finally,we realize the personalized recommendation of online learning resources.This paper verifies that the improved hybrid algorithm is more effective than traditional User-CF and association rule recommendation algorithm in terms of precision,F-measure,recall rate and average error Based on the results of experimental result.The main research work is as follows:(1)This paper analyzes the development history of online learning platform and puts forward the urgency and maneuverability of personalized recommendation in online learning.(2)It has been studied in the relevant technologies of the individualized algorithm.We introduce four commonly used recommendation algorithms and analyze their advantages and disadvantages,set the foundation for later research.(3)Article improves the hybrid algorithm by combining the knowledge unit closures found by association rules and the collaborative filtering based on LDA learning path analysis.The algorithm formed clusters and completed the mixed personalized recommendation based on the study path topic distribution analysis which is combined with the time dimension to obtain the student similarity and the knowledge unit closure.(4)By design of the comparison test,we verify that the hybrid algorithm is superior to the traditional recommended algorithm in terms of recall,accuracy,F-measure and MAE.(5)In the end,we summarize the current work.In view of the current algorithm's insufficient,we discuss and prospect for the future work.