| Knowledge Tracing(KT)is an essential component of intelligent online education and is the key to personalized instruction of students’ learning process.It aims at dynamically estimate students’ knowledge status and predict whether they will answer the next question correctly based on their historical answer records.Predicting the state of students’ knowledge is a difficult task because student learning is a dynamic process and the state of students’ knowledge changes.Existing approaches tend to ignore the fact that students’ learning ability developments in stages and do not take into account the influence of students’ short-term knowledge state on the next stage.Existing models do not emphasize the importance of students’ long-term knowledge state and short-term knowledge state,and the importance of information about both questions and concepts in many-to-many relationships is not well explored.From a modeling perspective,it is difficult for previous models to capture which information between question concepts better affects the probability of answering unknown questions correctly.To address the above two problems of current knowledge tracing models,this paper proposes two new knowledge tracing models based on the attention mechanism and graph attention neural network to solve the above two problems respectively,and designs an intelligent learning and analysis system based on knowledge tracing algorithms.(1)To address the problem that existing models neglect the importance of long-term knowledge states and short-term knowledge states,this paper proposes the Long-and Short-term Attention Network for Knowledge Tracing(LSAKT)model.Firstly,in this paper,the overall sequence is divided into time-stamp based subsequences,the first attentional neural network is designed to capture the long-term knowledge state of the learning student interacting with the performance of the problem,while another attentional network perceives the short-term knowledge state of the learning student according to the last subsequence.Subsequently,this paper combines students’ long-term knowledge states and short-term knowledge states to model students’ final knowledge states.The proposed model is finally evaluated in this paper on four publicly available real data sets,and the experimental results show that the LSAKT approach achieves optimal results when compared with other baseline models.(2)To address the problem that existing models have difficulty in capturing higher-order association information between problems and concepts.In this paper,we introduce Graph Attention Networks(GAT)and propose Graph Attention Embedding Dynamic Key-Value Memory Network for Knowledge Tracing(GDKVMN).Specifically,GDKVMN uses graph attention neural networks to aggregate feature information between problem concepts in a correlation-weighted manner,enriching the model’s input representation embedding information,while further taking into account the student’s learning rate and problem-solving ability factors in representing the learning process of the plug-in.Extensive experimental results on three real public datasets show that the proposed model in this paper achieves the best performance compared to the baseline model.(3)In this paper,we design an intelligent learning and analysis system based on knowledge tracing algorithm,which uses a knowledge tracing algorithm to track students’ knowledge level of practice problems and visualize and analyze students’ historical practice results.The significance of this system is that by analyzing students’ knowledge status,it visually shows students’ current weak knowledge points,which facilitates teachers to intervene in students’ learning in a timely manner,guiding students’ learning,trying to complement students’ weak concepts and related knowledge,providing personalized supplementary education,and improving students’ learning efficiency. |
