Research On Adaptive Learning Path Recommenda-Tion Method Based On Knowledge Graph

At present,China's "education+platform" information service mode is in a period of rapia development.In this process,learners' network resources are increasingly rich,and the teaching environment is gradually improving.However,at the same time,massive online resources bring learners "knowledge lost" and "information overload" and other problems.At present,most online education recommends learning paths to learners based on sequential knowledge structure,which separates the relationship between knowledge and leads to the mismatch between recommended resources and learners' needs.As a graph based data structure,knowledge graph can accurately represent the complex relationship between knowledge.In view of this,this thesis studies a learning path recommendation method based on the structured expression of junior high school mathematics knowledge by using knowledge graph and combining with learners'individual characteristics to give them adaptive adjustment,which is conducive to stimulate learners' interest in learning and improve learners' learning efficiency.The specific research contents are as follows:Firstly,according to the description of learners' personalized characteristics,this thesis analyzes the learning path recommendation method and learners' needs,divides learners into four characteristics:basic information,cognitive level,learning ability and learning style,and refines each characteristic,so as to build a learner model.At the same time,the improved Deterministic inputs,noisy "and" gate model model,the explicit and implicit Felder Silverman scale and the quantitative model of comprehensive learning ability designed in this thesis are used to obtain the characteristic values of learners.The quantitative model of comprehensive learning ability calculates learners' ability value from four dimensions,which are divided into knowledge learning efficiency,communication and cooperation ability,information acquisition ability and self-control ability,and uses data to prove the effectiveness of each dimension quantitative formula.Secondly,through the analysis of the knowledge structure of junior high school mathematics and the suggestions of domain experts,a three-layer knowledge graph structure was designed.In order to ensure the high quality and error-freeness of the knowledge graph,a combination of semi-automatic and manual methods was used to construct 517 nodes of junior high school mathematics.In the knowledge graph,the first-level knowledge points are called knowledge blocks,the second-level knowledge points are called knowledge chapters,and the third-level knowledge points are called knowledge points.At the same time,in order to better describe the importance of knowledge and the degree of dependence between knowledge,this thesis expands the third level of the knowledge graph,and adds improved node centrality and inter-node probability to it.Distribution makes the expanded knowledge graph more fine-grained to characterize the relationship between knowledge,and better data support for adaptive learning path recommendation.Finally,the learner model is combined with the expanded knowledge graph,and based on the learner's cognitive level,it is judged whether the learner is to perform the compensatory learning of junction tree reasoning or the continuous learning of node centrality ranking,and then according to the learner The learning ability and learning style screen the carrier of knowledge points in the learning path,and push the learning path and supporting learning resources to the learners.Based on the analysis of the experimental results,the learning path recommended in this thesis is reasonable and meets the individual needs of learners.And in the whole reasoning process,in order to improve the reasoning efficiency of the junction tree,it has been improved in the reasoning process.This thesis designs an adaptive learning guidance system around the adaptive learning path recommendation model,and divides the system into seven modules according to the five-in-one design concept of testing,learning,practice,testing,and assistance,and then realizes and verifies through these seven modules The effectiveness and rationality of the recommended model in this thesis are discussed.