Research And Implementation Of University Curriculum Recommendation System Based On Portrait

With the impact of the epidemic,China’s online education industry has developed rapidly.However,behind the rapid development,due to the lack of online education resource recommendation platform for college courses and the low accuracy of traditional recommendation algorithm,it is difficult to ensure the quality of online courses while providing resources matching students’ learning ability.Therefore,this thesis proposes an objective and comprehensive curriculum portrait system to recommend courses for college students.The curriculum portrait system proposed in this thesis is composed of five dimensions including curriculum characteristics,curriculum quality,curriculum difficulty,curriculum knowledge coverage and curriculum basic information.In this thesis,a non-random probabilistic dropout method is designed to help the FastText algorithm find the effective keywords except for the strong feature words,and thus solve the problem of overfitting in the process of classifying the text data of the course introduction text to obtain the course feature indicators.Bert model is used to analyze the sentiment of the review information and extract the course quality index.The Kmeans algorithm was used to cluster the students’ practice situation,and the course difficulty index was obtained.Jaccard similarity is used to cluster the course names,extract the course outline catalog to obtain the collection of each type of course knowledge points,and calculate the coverage index of course knowledge points.The basic information of the course is mapped into four-dimensional vector according to the rules,and the basic information index of the course is obtained.Through the statistical analysis of the portrait information of students’ course selection and collection and the data of students’ practice after class,the learner portrait is obtained.Based on the completion of the construction of the curriculum and student portrait,this thesis puts forward two recommended methods.The first is a recommendation method based on the similarity of the course portrait.When the user clicks the course,the course with the highest similarity coefficient with the course portrait is selected for recommendation.The second is a recommendation method based on deep confidence network,which combines the features of students’ curriculum portraits as the input of deep confidence network,and selects top-N courses for recommendation according to the model prediction score.This thesis uses B/S architecture for development,uses Scala language and SSM framework to realize the back-end business process,uses Vue framework for front-end development,uses relational database MySQL for data reading and storage,realizes the university curriculum recommendation system based on portrait,and designs test cases to complete the system test.By using real data sets and comparing with other popular algorithms,this thesis proves that the improved FastText model significantly improves the classification evaluation index,verifies the effectiveness of the nonrandom probability uptext method in alleviating the overfitting problem of the classification model,and also proves that the recommended method proposed in this thesis is expected to provide reference and help for college education.