Research On Deep Knowledge Tracing Model Integrating Knowledge Structure Graph

With the advancement of computer technology and the emergence of the period of education informatization,online education has emerged as one of the learning ways for students to master knowledge and enhance skills.In the teaching process of online education,if we want to improve the learning effect of learners,we need to provide personalized and intelligent teaching services for learners and realize the targeted tutoring for each learner.Therefore,how to scientifically assess learners’ knowledge mastery level based on their historical learning records and develop personalized learning programs has become a hot issue to be solved in online education.The purpose of Knowledge Tracing(KT)is to obtain learners’ knowledge mastery level and predict their future learning based on learners’ historical response records.Knowledge tracing is important for personalizing online learning.At present,researchers have proposed new methods and models in the field of knowledge tracking to combine information such as learner features and domain features with knowledge tracking,and have achieved certain results.However,there are still some problems: the structural relationships of knowledge points are not rich enough when solving the fusion problem of domain features;the embedding representation of multiple relationships between knowledge points and knowledge points needs to be improved.Therefore,this study makes an attempt in solving the above two problems and proposes a deep knowledge tracking model(Tr-dkt)with fused knowledge structure graphs,and the main research work is as follows:(a)Constructing a knowledge structure map,Based on learners’ historical learning records and the pyramidal theory of knowledge organization,this study investigates the potential links between isolated knowledge pieces,and constructs five types of knowledge relationships including antecedent-successor,similarity,parent-child,sibling and parallel relationships.(b)Representation of the knowledge structure map,this study uses the Translating Embedding algorithm to embed the knowledge structure map to obtain a vector representation of knowledge.The characterized knowledge structure map is fed into the knowledge tracking model together with the learners’ historical answer records to accomplish the goal of integrating knowledge tracking into the knowledge domain feature domain.(c)Experimental analysis,this study investigates the influence of the number of knowledge point relationships on the deep knowledge tracking model through experiments,analyzes the degree of influence of antecedent-successor,similarity,parent-child,sibling and parallel relationships on learners’ knowledge mastery level,discusses the trend of the influence of the embedding dimension of the knowledge structure map on the performance of the knowledge tracking model,and compares it with other knowledge tracking models,and obtains a more satisfactory The experimental results are more satisfactory.