Research And Implementation Of Personalized Exercise Recommendation System Based On Knowledge Tracing

With the development of internet technology,online learning has become a common way of learning.However,in online learning,teachers find it difficult to capture students’ learning reactions and knowledge mastery status,and provide targeted suggestions.At the same time,facing massive teaching resources,users also find it difficult to quickly and accurately find course exercises that are suitable for their learning progress and knowledge mastery level.Common methods for evaluating students’ knowledge mastery include cognitive diagnostic models and knowledge tracking models.Evaluate students’ knowledge mastery status by analyzing and mining their historical answer data.In the current personalized exercise recommendation field,there are mainly methods such as using deep learning methods to model users and using collaborative filtering to recommend.In common knowledge modeling methods,most models are used to automatically extract the features of exercises and knowledge points,ignoring the complex contextual knowledge background and mapping relationship between exercises and knowledge points during the student learning process,and not considering the group effect during the student learning process.The problem recommendation strategy based on collaborative filtering does not fully consider students’ own learning characteristics and knowledge mastery,resulting in inaccurate recommendation results.In this thesis,a personalized exercise recommendation strategy based on knowledge tracking is proposed to address the above issues.Firstly,use the corresponding relationship between exercises and knowledge points in the Q matrix and students’ performance in answering knowledge points to obtain the Q matrix of upper and lower cultures.Using the extracted feature vectors of knowledge points and students’ answers as inputs to the deep knowledge tracking model,a more accurate understanding of students’ knowledge mastery status can be obtained.To predict the probability of students correctly answering exercises and calculate the personalized"difficulty" coefficient of exercises for students.And use neural networks to predict the probability vector of knowledge points that students need to learn in the next time period from the interactive sequence of students’ historical problem-solving,and obtain the students’ learning range.Then the collaborative filtering algorithm is used to find similar student groups for users,and the group effect is used to calculate and update students’ knowledge point mastery status and the range of knowledge points to be learned in the next time step.Taking into account students’ level of knowledge mastery and the range of knowledge points to be learned,recommend exercises with appropriate difficulty and knowledge point coverage that are in line with their learning progress.And re filter the recommended list to remove exercises with high similarity in the set of exercises and improve the diversity of recommendation results.Finally,through requirement analysis of the system,a personalized exercise recommendation system was designed and developed.It mainly includes the teacher end and the student end.The student end has functions such as student exams,error books,and recommendation of test questions.The teacher end has functions such as exam management and exercise management.Adopting a B/S architecture that separates the front-end and back-end,mainly using technologies such as SpringBoot and Vue.js for development and design.And the completeness and stability of the system were tested.