Research On Knowledge Tracing Mechanism In Online Learning Platforms

With the rapid development of e-learning,more and more students and individuals actively participate in the e-learning platforms to learn new skills in order to cope with the rapid development of society.However,in order to achieve the efficient and independent learning in the e-learning environment for students,and implement the intelligent and personalized teaching for e-learning platforms,how to model learning process of students and accurately track the learning knowledge states of students are challenges that researchers must resolve.Although many knowledge tracing methods have been proposed at present,there are still some shortcomings in dealing with the knowledge tracing problem in the real e-learning environment.These shortcomings are mainly reflected in the following aspects:(1)When the current methods analyze the learning data,the complexity of student learning data is not fully considered,so the main features of student learning data cannot be explored.(2)When the current methods solve the problem of knowledge tracing,they are mainly from the perspective of solving mathematical problems.So these models are poorly interpretable and the accuracy of knowledge tracing is not high enough;(3)Most knowledge tracing methods are based on the student's personal learning history data to trace the student's current knowledge state,so the student's new knowledge learning process cannot be predicted;(4)The current knowledge tracing methods mainly focus on how to improve the accuracy of knowledge tracing,without thinking about the training and update mechanism of the model in a real e-learning environment.In this paper,in order to solve the existed problems of the current knowledge tracing methods in the real e-learning environment,the research on the knowledge tracing mechanism in the online learning platform is mainly divided into the following parts:(1)This paper observes the students' learning data firstly and then proposes the stacked autoencoder based on long short-term memory network method to mine the temporal features of students' learning,and defines the students' learning ability features.Last,we cluster students based on these two features and analyze their learning data.(2)According to humans learning and the cognitive process,the paper improves the current deep knowledge tracing model and proposes knowledge tracing model based on learning and memory process,which enhances the interpretability of the model and improves the knowledge tracing effect.(3)According to knowledge tracing model based on learning and memory process and the clustering features of students' learning data,this paper proposes a new knowledge tracing model based on group learning features,which can accurately predict the learning process of new knowledge for students.(4)According to the clustering features of students' learning,this paper firstly proposes a training method for knowledge tracing model based on learning and memory process,which enhances the individualized ability of the model.Secondly,this paper optimizes the training process for new knowledge tracing model based on the group learning features,so that it can effectively deal with the knowledge tracing problem in the real e-learning environment.Finally,the experimental results on two real datasets demonstrate that the knowledge tracing the effect of knowledge tracing model based on learning and memory process is better than the current knowledge tracing method,which verifies the feasibility of the new knowledge tracing model based on the group learning features.In addition,the paper analyzes the impact of the clustering features of students learning on the knowledge tracing effect through experiments and proves the importance of the clustering features of students learning for knowledge tracing.