Research On Application Of Behavior Analysis And Recommendation System For Personalized Learning

The development of online education has brought massive data on learning behavior,promoted the development of personalized learning,and greatly expanded the spatial and temporal boundaries of teaching.New students who have just entered the university need to establish new learning models and social relationships,and colleges also encourage students to dig deeper into their interests and innovate independently.In today’s higher education,students can use online platforms to assist learning anytime,anywhere,and on demand,use instant messaging platforms to discuss issues with teachers and students,and log on to MOOC platforms for additional teaching resources.Rich online education methods can meet the personalized learning needs of students at different stages of higher education,but in practice,they still face the problem of insufficient learning quality and efficiency.In order to solve these problems,This paper analyzes and studies two typical application scenarios in personalized learning of online education under the background of higher education,namely hybrid learning and MOOC learning,and draws a series of valuable conclusions and solutions to their respective problems.Students follow a unified curriculum plan in the classroom,and use the platform to achieve blended learning that combines online and offline.In this scenario,a considerable part of the learning behavior occurs online,so teachers are difficult to grasp the online learning of students,and students can’t get timely feedback.Both sides can only grope according to past experience,and urgently need a lot of theoretical support for teachers and students to dynamically formulate personalized learning strategies,and effectively improve the teaching quality.This paper uses data mining technology to study the multi-platform learning process,learning mode,learning influencing factors and platform characteristics of students on instant messaging platforms and online learning platforms in view of the lack of understanding of online learning processes by teachers and students in blended learning,which filling in the relevant research gaps.MOOC learning can help students obtain multi-angle,multi-level quality educational resources beyond curriculum planning,and support students to carry out lifelong learning outside the classroom,but it also makes students drown in a large amount of information.It is difficult for students to quickly match learning objectives and teaching content,which means reducing learning efficiency.The recommendation system can solve this problem,and the existing recommendation system applied to the MOOC platform usually assumes that the interaction behavior of the learner and each video has the same status,and does not dig into the information contained in the fine-grained viewing behavior,so that the noise generated by the trial behavior in the target matching process is introduced,and the system will also ignore the difficult content that the learner repeatedly studies.In addition,MOOC learning has a low binding force and high degree of freedom,allowing learners to have multiple learning objectives.But the existing MOOC recommendation system ignores this possibility and only models a single user vector.Therefore,this paper proposes a multi-objective recommendation algorithm based on fine-grained behavior sequence,FSMRec,which effectively improves the model recommendation performance and module interpretability by designing the video importance formula for mining learning target directivity information in fine-grained behavior,and introduces a multiobjective extraction module that captures personalized learning goals.So that FSMRec can providing learners with more efficient video recommendation services.We have proved the reliability and superiority of the proposed model through a large number of experiments and case studies.