Research On Online Question System For Personalized Learning

With the popularization of Internet and computer technology,online learning platforms have developed rapidly.The online learning platforms provide students with convenient learning conditions and generate a large amount of learning behavior data.Mining the massive students’ behavior data helps to gain a deeper understanding of their learning situation,thus providing personalized learning resources for students and improving their learning efficiency.Knowledge Tracing,which realizes personalized modeling by analyzing student behavior sequences,has become a research hotspot in the field of educational data mining.Currently,although the deep learning based knowledge tracing method has achieved certain results,there are still problems such as insufficient utilization of students’ multiple learning behaviors and insufficient research on complex learning behaviors.Meanwhile,although there are a lot and various kinds of online learning platforms,they lack efficient aiding for student’s second half of the learning process.Aiming the above problems,this paper focuses on online learning scenario and targets personalized learning,carries out online question system research and data mining based on knowledge tracing.The contributions of this paper are as follows:1.Design and develop an online question system integrating knowledge tracing algorithms to assist students in personalized exercises and exams.We carry out function testing and stress testing to assure the system’ reliability;2.Propose a multiple learning behavior data standard and DKT with Multiple Learning Behavior(DKT-MLB)model.This model uses exercise behavior as a reference behavior,and uses knowledge concepts as links to joint other learning behaviors.By using this strategy,the proposed model can fully utilize the multiple learning behaviors generated in the online learning system,and more accurately restore the learning process of students,solve the problem of existing knowledge tracing models limited to exercise behavior;3.Propose Programming Debugging Ability Tracing(PDAT)to model student’s programming debugging ability.This model formulates student’ programming debugging behavior based on knowledge tracing definitions.For the annotation task of code changes generated by debugging behavior,the CodeBERT-PD model is introduced,and domain knowledge is injected into the original CodeBERT model through a small number of sample fine-tuning methods to ensure the accuracy of model annotation and improve the efficiency of annotation.At the same time,we use DKT model to tackle annotation sequences,extend knowledge tracing to the complex programming debugging scenario,achieve the tracing of students’ programming debugging ability.The above systems and algorithms have been applied in the course teaching of Beijing University of Posts and Telecommunications.Based on the learning behavior data generated from real teaching processes,the proposed model was compared with multiple baseline models,which can verify the effectiveness of the proposed model.