Online Programming Exercises Based On The Knowledge Graph Recommend Visual Analytics Methods

Online programming platforms have the characteristics of convenient learning pathways,diverse learning times,and abundant teaching resources,and are gradually becoming an important learning method.Faced with a massive amount of learning resources,learners generally face problems such as "information overload" and "knowledge loss",resulting in online programming platforms being unable to fully utilize the role of intelligent tutoring.Therefore,how to provide accurate and effective personalized programming exercise recommendations based on learners’ own characteristics is one of the urgent problems to be solved in online learning.Most of the existing programming exercise recommendation work uses methods based on collaborative filtering and cognitive diagnosis,but both of them have shortcomings in the quality of exercise recommendation.On the one hand,recommendation methods based on collaborative filtering complete the recommendation by mining the potential interest preferences of learners.However,such methods only rely on the history of question making records to calculate the similarity of questions in the recommendation process,ignoring the attribute information of the exercises themselves,which makes it difficult to comprehensively capture the similarity of exercises.On the other hand,the recommendation method centered on cognitive diagnosis aims to help learners acquire more knowledge,first modeling the learners and then recommending exercises that match their cognitive level.However,such methods are usually based on predefined/single recommendation rules,making it difficult to adapt to the complex and diverse needs of learners and constantly changing learning objectives.At the same time,learners often can only obtain a list of recommended exercises through intelligent algorithms,without knowing the basis for their recommendation,which hinders learners’ choice and trust in exercises.In order to solve the above problems,this thesis proposes a visual analysis method for online programming exercise recommendation based on knowledge graph.Multiple exercise recommendation strategies are designed for learners with different cognitive levels and learning objectives,and visual analysis technology is introduced to demonstrate the exercise recommendation process for learners.The main research content of this thesis is as follows:1.This thesis proposes an online programming exercise recommendation method that integrates knowledge graphs.The exercise number,knowledge point and difficulty information are used as entities to build a knowledge graph,and the improved knowledge representation learning model is used to vectorization the exercise.On this basis,the exercise vector is added to the recommendation process of collaborative filtering to make up for the shortcomings of existing recommendation algorithms,so as to obtain more accurate exercise recommendation results.2.This thesis proposes a weak knowledge point recognition method based on a neural-cognitive diagnostic model.Modeling learners’ cognitive level based on neural-cognitive diagnostic model,predict the probability of learners correctly answering the online programming exercises based on their practice records,and obtain a representation vector of learners’ knowledge mastery through the neural network interaction layer during this process.3.This thesis constructs an interactive exercise recommendation visual analysis system.Based on learners’ cognitive level differences,define diverse exercise recommendation strategies based on their different learning needs in terms of knowledge points,difficulty,and pass rates.We have designed and implemented an interactive visual analysis system that supports learners to complete exercise recommendation exploration from overall overview to detailed exploration,as well as multi view linkage.We provide an overview of learners’ problem-solving behavior and cognitive level comparison,and use rich visual metaphors to display exercise recommendation results,helping learners intuitively perceive the exercise recommendation process and basis.This thesis uses an exercise submission dataset from an online programming platform for experiments,and proves the rationality and effectiveness of the method and system in analyzing learners’ learning situation and recommending online programming exercises through multiple cases.