Research On Personalized Learning Method Based On Knowledge Graph

With the continuous development of information technology and modern educational theory,personalized learning has become a hot spot of common concern.The fundamental purpose of personalized learning is to take appropriate measures to fully meet the individual needs of learners according to the personality characteristics of learners,so that the learners’ personality can be fully developed.The core content of personalized learning mainly includes learning diagnosis and learning recommendation.(1)Learning diagnosis: analyzing the learner’s learning level and learning ability by analyzing the learner’s personality characteristics and learning data;(2)Learning resource recommendation: according to the learner’s learning level and ability,recommend the learning resources that are most suitable for learners.The current learning diagnosis method was staying at the level of learner’s answer to the questions,but did not consider the mastery of several knowledge points that make up the questions,thus lead to insufficient diagnosis depth and poor diagnostic results.At the same time,the learning resource recommendation methods mostly recommend similar resources according to the learner’s historical learning interest,and do not take the learners’ learning defects into account and can not recommend learning resources that learners really need.In view of the above problems,This paper studied the diagnosis method for learner’s learning level from the level of knowledge points and the recommendation method of learning resources according to the learner’s learning defects.The research content mainly included the following two aspects:(1)Compared the three theoretical foundations of the existing learning diagnosis methods and analyzing their existing deficiencies,the techniques of knowledge representation and knowledge inference were studied and applied to the storage and representation of the relationship between knowledge points,and a personalized learning diagnosis method based on knowledge graph was proposed.Firstly,constructed the knowledge graph of the subject.Then,designed the matrix evolution algorithm,calculated the learner’s mastery of the knowledge points through the test results,and obtained the rough knowledge state of learners.Finally,according to the knowledge graph,the knowledge inference technology was applied to infer the learner’s rough knowledge state,and the learner’s final knowledge state was obtained,so as to explore the root of the learner’s knowledge defect and provided data support for the learning resource recommendation.(2)Compared several theoretical foundations of existing learning resource recommendation methods and analyzes their existing deficiencies.Based on the learner’s learning defects,a personalized learning resource recommendation method based on knowledge state was proposed.Firstly,according to the learner’s knowledge state obtained by the learning diagnosis method,the unmastered knowledge points in the knowledge state were regarded as the learner’s weak knowledge points.Then,the subject knowledge graph and the weak knowledge point set were combined,and the topological sorting was used to generate the learning points to be learned.After that,the similarity between the sequence of knowledge points to be learned and the knowledge points covered by the resources in the learning resource pool was calculated by the Jaccard similarity algorithm.Finally,the learning resource sequence to be recommended was generated.Thereby,it is realized to fundamentally recommend learning resources for learners.(3)Experiments showed that the knowledge learning method based on knowledge graph can obtain the learner’s learning defects more accurately.And in view of this,recommended the learning resources that the learners really need.In the experiment,200 students in a second grade were tested.The average diagnostic accuracy of the students’ knowledge level was up to 81.13%.After the students recommended the resources,the average score was compared with the students who did not recommend.The increase was 15.7% higher.Therefore,the proposed method had good diagnostic and recommendation performance,which provided a new and effective learner knowledge level assessment and learning resource push strategy for personalized learning.