Research On Learning Path Recommendation Algorithm Based On Course Knowledge Graph

With the rapid development of "Internet plus education" and the increasing abundance of educational resources,the importance of learning path planning is gradually growing owing to the high difficulty and time consuming nature of finding a suitable learning path for learners in the face of massive learning resources.The recommendation algorithm for learning path recommendation,based on learners’ preference information and educational principles,has become one of the research hotspots in the field of recommendation algorithms.In this paper,we study the learning path recommendation algorithm based on knowledge graph,and the main work is as follows:(1)Building a course knowledge graph model.The paper takes the data structure course as the research object,takes the knowledge points as the center and the learning resources as the supplement,uses BERT+Bi LSTM+CRF to obtain the word vector with richer semantic information and conducts the course knowledge point entity extraction,further uses the way of template rule to extract the five semantic relations among the course knowledge points to ensure the accuracy of the relations among the knowledge points in the course knowledge map,and uses Neo4 j graph database is used to store the data.(2)A knowledge point recommendation algorithm which fuses temporal information and knowledge graph neighborhood double-ended information is proposed.At first,the preference propagation and neighbor information aggregation are performed at the learner end and the knowledge point end alternately using the knowledge graph for sharing the latest information obtained from each other and improving the quality of feature extraction at both ends.The temporal sequence information between learners and knowledge points is enhanced.Then,the sequence information and global features are obtained using Bi GRU and the self-attention mechanism respectively,thus obtaining the learner feature vector with temporal attributes.Finally,the interaction probability is calculated for both ends of the feature vector.The experiment result shows,compared with the optimal baseline model,the proposed algorithm improves the AUC and ACC by 1.29% and 1.66%,respectively.(3)A learning path recommendation algorithm based on the course knowledge graph is put forward.Firstly,the learner’s information is modeled from multiple dimensions to obtain the personalized characteristics of the learner.The learning objectives are obtained using the CTIKGND algorithm and mapped accordingly.Then,the global and sequential learning path recommendation strategies based on the course knowledge graph are proposed respectively to plan the optimal learning paths according to the learners’ learning styles.Finally,the learning path recommendation system is designed and developed,and the course knowledge graph query,personalized learning path recommendation and knowledge point query functions are tested to provide comprehensive learning support.In summary,the paper firstly investigates the construction of course knowledge graph model,then proposes the CTIKGND algorithm to realize the prediction of learners’ learning goals,and finally constructs a learning path recommendation algorithm based on course knowledge graph to plan learning paths for learners,and designs and implements a learning path recommendation system accordingly.