Research And Practice Of Recommendation System For JavaWeb Online Learning Resources

Flipped classroom teaching is suitable for computer programming courses.For some advanced programming courses,Viewing and learning high-quality online learning resources can be an important means for learners to improve their autonomous learning effect before flipped classroom classes.This paper takes Java Web online learning resources as an example to study how to solve the problem that teachers can not manually screen and recommend high-quality online learning resources for learners in the case of diversity and complexity of online learning resources and learners’ learning interests may change in the learning process.The use of automated recommendation technology has become a better solution,but through the research on the status quo of recommendation technology at home and abroad,it is found that there is still a dispute in the current industry about which recommendation method has a better effect.Therefore,this paper studies and practices how to effectively combine Java Web online resources with recommendation technology to achieve better recommendation.In view of the above difficulties in recommending high-quality resources,this paper has studied and analyzed the relevant theories and technologies of the recommendation system,and selected a hybrid recommendation method with better recommendation performance to design the system.According to the requirements of hybrid recommendation methods,based on the characteristics of learners and Java Web online learning resources,the basic recommendation methods are designed from four perspectives: the content of Java Web online learning resources,learners’ scoring behavior,changes in learners’ learning needs,and the design of recommendation systems.These recommendation methods include content-based recommendation characterized by knowledge points of learning resources,model-based collaborative filtering recommendation characterized by learner ratings,real-time recommendation based on changes in learners’ learning needs,recommendation based on the number of learner ratings,recommendation based on the release time of learning resources,and high score resource recommendation based on learner preference knowledge points to solve the cold-start problem of the recommendation system.Based on the basic recommendation method,this paper designed recommendation modules based on the parallel hybrid recommendation method and recommendation list.For the recommendation function module,this paper designs a learner model and a network learning resource model,and designs recommendation algorithms based on these two models,specifically: In the integrated recommendation module,the LFM based collaborative filtering recommendation algorithm is adopted;In the personalized recommendation module,real-time recommendation algorithm is used to generate recommendation results based on learners’ historical scores and current scores of resources,and then mixed with high scoring resources in learners’ preferred knowledge points for recommendation;In the module of similar resource recommendation,a content-based recommendation algorithm with learning resource knowledge points as labels is adopted;In the resource recommendation module with high rating rate,a recommendation algorithm based on the number of learners’ ratings is adopted;In the latest release resource recommendation module,the recommendation algorithm based on the time of release of learning resources is adopted.In the specific development of the system,the Linux operating system is used as the service running environment,Spark is used to implement the algorithm,Flume,Kafka and Kafka Stream are used to filter,process and send the learner scoring information in the log file in the real-time recommendation,and Mongo DB and Redis are used to develop the database.The use of these technologies is conducive to the improvement and expansion of the system in the future.In the evaluation and analysis of the system,online experiments,user surveys and learner interviews are used to evaluate the system.The online experiment adopts the experimental evaluation based on Top N recommendation.By calculating and comparing the recall rate,accuracy rate and F1 value of mixed recommendation,LFM based recommendation and scoring based recommendation with different recommendation list lengths,this paper determines the length of the recommendation list when the system recommendation effect is good.The user survey adopts a questionnaire design based on Likert scale,which mainly includes learners’ evaluation of the system’s interface usability,resource recommendation satisfaction and the overall system.Through the analysis of the questionnaire results,learners have a high degree of recognition of the system,and believe that the hybrid recommendation method adopted by the system is better than the LFM based recommendation and the rating rate based recommendation method.Learner interviews are conducted for the usefulness,usability and satisfaction of the system.According to the analysis of the interview results,six interviewees relatively recognized the usefulness and usability of the system,and most of them were satisfied with the system.In general,this system can better recommend Java Web online learning resources,so this research has certain research and practical application value.