| Personalized teaching and learning is an important topic in educational research.The rapid development of the Internet has pushed personalized learning to a new high level,and provided abundant learning resources for people.However,the explosion of learning resources also brings the burden of choice on people.As an important learning resource,exercises play an important role in testing learners’ learning outcomes and determining their mastery of knowledge,and are an integral part of personalized learning.Exercise recommendation technology is an important solution to the problem of overload of learning resources,meets the individual needs of learners.It is an important means to enhance the efficiency of learning for learners,and has become an important topic in the field of personalized learning.Currently,there are two problems in personalized exercise recommendations,i.e.,sparse data and neglect of semantic information about knowledge points.To solve the problem of sparse data,this thesis investigates an exercise recommendation method that combines deep knowledge tracking with knowledge matrix complementation.To solve the problem of ignoring semantic information of knowledge points,we first present the construction of mathematical subject knowledge graph,which contains rich semantic information.Then,we investigated a method for recommending exercises based on the knowledge graph.Contributions of this thesis are as follows.1.This thesis proposes an exercise recommendation method that combines deep knowledge tracking and knowledge complementation to alleviate the problem of data sparsity in exercise recommendation.Learners’ knowledge levels are modeled by deep knowledge tracking,while incorporating learning information from similar groups.Further,the SVD++ method is used to supplement the learner knowledge level matrix to alleviate the problem of data sparsity.Experimental results show that the method has significant advantages in accuracy,recall and F1 value.2.This thesis presents the construction of a mathematical subject knowledge graph.First,based on the Assistments2009 public dataset,the answer records with knowledge points in the original data were retained through data cleaning.Secondly,the knowledge graph entities were extracted,and the relationships were designed by experts to construct the knowledge graph of mathematical subjects.Finally,Neo4 j is used to store the knowledge graph and realize the visual representation of the knowledge graph.3.This thesis proposes a knowledge graph-based exercise recommendation method.Firstly,the learner’s knowledge level matrix is obtained using a dynamic key-value memory network.Secondly,the relationships between the knowledge graph nodes are given weights by expert evaluation of the closeness of the relationships between the knowledge points.Finally,based on the learners’ knowledge level matrix and the relationship weights between nodes,the knowledge level values in the nodes of the knowledge graph are supplemented,and the completed knowledge graph is used to recommend exercises to learners.This method utilizes semantic information between knowledge points to alleviate the problem of data sparsity.Experimental results show that this method has more advantages in precision,recall,and F1 values. |
