Research On Key Technologies Of Personalized Learning Path Recommendation Based On Knowledge Graph

Online learning mode,with its universality,popularity and intellectualization,is becoming the key field of intelligence education.The number and scale of online learning platforms and network learning resources are increasingly augmented,and the group of self-directed learners is gradually expanding.Due to the lack of professional learning planning and guidance,learners face the dilemma of "information overload" and "learning loss" caused by massive learning resources,resulting in chaotic learning process and low learning efficiency which can be effectively solved by personalized learning path recommendation technology.This thesis designs A Polybasic Personalized Learning Path Recommendation Model Based on Course Knowledge Graph(CKG-2P LPR)for online learners who are concerned about job-seeking and self-improvement.By completing the design and construction of course knowledge graph,polybasic learner portrait and personalized learning path recommendation algorithm,CKG-2P LPR can help learners achieve their learning objectives efficiently.The main research work of this thesis is as follows:1)A course knowledge graph based on online course data is built.By taking “professional skills-knowledge points-online courses” as the graph framework,mining and defining the knowledge points and various attribute information contained in the course,this thesis semi automatically constructs the course knowledge graph in the field of computer software which supports the design and implementation of subsequent personalized learning path recommendation algorithm.2)Based on online learning behavior data,a polybasic learner portrait is designed.Firstly,a professional skill label tree is constructed to quickly determine learning goals;Secondly,a technology of self-Learning Tags Deep Knowledge Tracing with Hybrid-Responses(LTDKT-HR)is proposed to evaluate learners’ knowledge mastery which solves the problems of inaccurate prior knowledge evaluation and repeated learning of knowledge points;Thirdly,a learning behavior feature model is constructed to represent learning preferences and a learning behavior feature scale is formulated for beginners to solve the problem of user cold start.Then,a polybasic learner portrait update mechanism is developed to support the dynamic adjustment of learning path based on learning progress.3)Combined with course knowledge graph and polybasic learner portrait,the personalized learning path recommendation algorithm is constructed.Firstly,an alternative course sequence generation algorithm based on knowledge graph is proposed to determine the learning content and course selection range;Secondly,a technology of Knowledge Graph Enhanced Recommendation with User Collaboration(KGER-UC)is designed which solves the multi-dimensional calculation problem of the recommendation degree of alternative courses,so as to provide a basis for measuring learning path recommendation;Thirdly,a personalized learning path generation algorithm and a dynamic updating mechanism of learning path are designed,which can not only recommend target preferred path to initial learners,but also recommend process preferred path to process learners according to their learning progress,thus improving the accuracy and matching of recommendation.4)The KGER-UC technology and CKG-2P LPR model proposed in this thesis is analyzed and evaluated by the comparative experiments.The experimental results show that the personalized learning path generated by the KGER-UC technology and CKG-2P LPR model can ensure the high quality of recommendation,thus better solve the problems of "information overload" and "learning loss" faced by learners.