Difficulty-based Knowledge Point Clustering Algorithm Using Multi-Dimensional Time Series Data Analysis

Online learners have different learning foundations,learning styles,and learning abilities,resulting in encountering different difficulty levels of knowledge points in courses.Teachers need to keep abreast of the difficulty level of knowledge points for learners of different cognitive levels in order to better tailor their teaching.Through collaborative analysis of multidimensional interactive behavior data generated by learners on the learning platform,two knowledge point difficulty classification algorithms are proposed as follows:1.To solve the problem that the existing knowledge point difficulty classification algorithms do not effectively consider learner interactive behavior patterns and forgetting behavior,this thesis proposes a knowledge point difficulty clustering algorithm based on multidimensional time-series data and learning path networks.First,the similarity of the difficulty between knowledge points based on student-system interactive behavior is inscribed by combining the group directed learning path network、the forgetting behavior of learners and student-system interaction degree.Second,the JMSD similarity model is improved by fusing the knowledge point popularity difference and the knowledge point average interaction degree difference to inscribe the similarity of the difficulty between knowledge points based on student-teacher and student-student interactive behavior.Finally,the knowledge point difficulty similarity matrix is obtained by integrating the difficulty similarity of knowledge points obtained from student-system interaction behavior,student-teacher interaction behavior and student-student interaction behavior.The spectral clustering algorithm achieves knowledge point difficulty classification based on the obtained similarity matrix.2.To address the problem that existing knowledge point difficulty classification algorithms do not consider the implicit learning patterns in learners’ learning behaviors,a knowledge point difficulty clustering algorithm based on multidimensional time-series data and maximum frequent subgraphs is proposed.Firstly,the algorithm constructs individual directed learning path graphs through student-system interactive behaviors,mines the maximum frequent subgraphs in the set of directed learning path graphs using the improved g Span algorithm,and inscribes the difficulty similarity of knowledge points based on student-system interactive behaviors by combining the maximum frequent subgraphs and student-system interaction degree.Inscribing the difficulty similarity of knowledge points based on student-teacher and student-student interaction behaviors according to the improved JMSD similarity metric model in the knowledge difficulty clustering algorithm based on multidimensional time-series data and learning path networks.Finally,three difficulty similarities of knowledge points are fused,and a spectral clustering algorithm based on the obtained similarity matrix is used to achieve knowledge point difficulty classification.The proposed algorithm is validated by multi-dimensional time-series data and test data of learners in GUANGXILIJIANGXUETANG.The experimental results show that it has a better effect of classifying the difficulty of knowledge points than the existing methods,which is helpful to assist teachers in continuously optimizing the design of teaching contents and teaching according to the material.