Research On Student Knowledge Tracing Algorithm Based On Deep Learning

The development of artificial intelligence and the rise of online education have ac-celerated the process of intelligent education,and knowledge tracing is one of the most basic and important tasks.It dynamically evaluates students’ knowledge mastery level according to their historical answers.At present,a lot of work has explored knowledge tracing,but there are still some issues that need to be addressed.For example,most of the existing methods do not consider students’ individual learning ability,rarely use global information when learning question representation,and the prediction of students’ an-swer performance needs to consider the long-term dependence in the sequence of history records.Aiming at the above problems,this paper conducts an in-depth study from two aspects of students and questions.The main work includes:1)A knowledge tracing method based on students’ learning ability is proposed.Individual learning ability is an important factor in the learning process,which affects students’ answer performance and knowledge absorption.The method proposed in this paper will model students’ knowledge state and learning ability at the same time,so that students’ learning process can be analyzed from two perspectives of knowledge and abil-ity.Firstly,the method calculates the students’ learning ability on specific knowledge concept according to the history records,and designs a sequential neural network to inte-grate learning ability as an important factor into the process of acquiring and updating the students’ knowledge state.Secondly,the encoder-decoder structure is used to extract the knowledge state related to the question to be predicted.Finally,the knowledge state and learning ability are combined to predict the students’ subsequent answer performance.2)A knowledge tracing method based on heterogeneous global graph neural net-work is proposed.In order to better learning question representation,this paper proposes a knowledge tracing method based on heterogeneous global graph neural network,which considers the global information in students’ question records,combines the question-knowledge concept inclusion relationship,the transfer relationship between questions and the knowledge concept co-occurrence relationship to construct the graph structure,and learns the question representation through the graph neural network algorithm.Finally,students’ knowledge state is modeled based on Long and Short-Term Memory.3)A knowledge tracing method combining historical state-aware is proposed.Since students’ current performance is influenced by their interaction with past questions,this paper proposes a knowledge tracing method that combines historical state-aware.From the perspective of the importance of historical knowledge state,this method divides students’ knowledge state into current knowledge state and historical important knowledge state,and obtains the current knowledge state through the Long and Short-Term Memory.Weighted aggregation based on the importance coefficient generates the historical impor-tant knowledge states.Finally,the two knowledge states are fused to predict the students’ subsequent answer performance.