Pattern Extraction And Visual Analysis Of Student Behavior Based On MOOC Data

In recent years,with the rapid development of intelligent technology and big data technology,Massive Open Online Courses(MOOCs)have rapidly emerged in countries around the world,attracting millions of users to learn and communicate online.MOOC data contains not only learner information and learning outcome data,but also web log records of learners’ interactions with various course materials,providing a basis for education stakeholders to gain insight into online learning behavior.However,student log records are characterized by large,complex heterogeneous and multiple data,and it is an important challenge to effectively extract and visualize the student learning patterns embedded in MOOC data.Visual analytics leverages human visual perception to analyze data through interactive exploration,engaging human cognitive abilities in large-scale,high-dimensional and sparse data analysis.Visual analytics can take advantage of human cognitive abilities to explore the internal patterns of data more effectively than traditional methods that rely solely on automatic computer analysis.Therefore,this paper proposes to apply advanced data mining and machine learning methods,combined with visual analytics,to process MOOC data in order to help teachers and education experts better discover patterns and laws of student behavior and understand the reasons behind various learning behaviors.Based on the heterogeneous and complex MOOC data,this paper proposes a approach based on data mining and visualization to extract student behavior patterns from different perspectives and help educational analysts and related researchers explore the potential patterns of student learning sequences in an interactive manner,providing new insights for subsequent curriculum improvement and targeted interventions.The main work of this paper is as follows.(1)Patterns extraction and visual analysis of student behavior based on multiattribute event sequences: this work models the collected learning record data as multi-attribute event sequences and proposes a multi-attribute event sequence pattern extraction based on Minimum Description Length(MDL)for the problem of different formats of attribute data and the existence of correlations between attributes.The method supports the adjustment of attribute weights and takes into account the correlation between attributes.Finally,an interactive visual analysis system SPVis,is designed to integrate rich views and multiple interaction methods to help users explore the learning patterns of different learning groups from different levels.(2)Visual analysis of student behavior and performance relationships based on whole-course clickstream data: This work uses higher-order networks to extract subsequences with dependencies in clickstream data,summarizes student learning behaviors based on the relationships between the subsequences,and models the operation process within videos.Meanwhile,this work designs four linked views,including pattern overview map,clickstream detail map,sequence view and clickstream comparison map,which not only can effectively overview a large amount of clickstream data,but also support detailed comparison of individual student behaviors,which helps users to deeply understand and analyze clickstream data.Finally,this paper conducts experiments based on real MOOC data,and demonstrates the effectiveness of the visual analytics system in helping teachers analyze learning behaviors,explore learning pattern,discover learning pattern and inspire course improvement through case studies and expert interviews.