SPOC online video learning clustering analysis: Identifying learners' group behavior characteristics

With the widespread of Small Private Online Courses (SPOC) in colleges and universities, course organizers who provide high-quality personalized course activities need to understand learners' learning status and characteristics, and then optimize the course teaching. However, little research has been done on different learners' group behavior characteristics, such as which indicators can reflect learner groups' behavior, and what are the typical behavior characteristics of different learner groups. In this work, we established a Python Language Foundation self-built SPOC course, and 109 undergraduates' learning behavior data were collected and analyzed. From 74-dimensional behavior log data consisting of 72 video-viewing, course video score, and comprehensive score, Principal Component Analysis was performed to reduce dimensionality. Agglomerative hierarchical clustering was used to divide learners into different categories, and the results showed that 15 groups are clustered. According to the analysis of the four indicators for group characteristics, which are the completion and viewing-stability of task-point videos, unit test excellence, and comprehensive score, it is identified into five primary types, including positive type, regular type, negative type, semi-negative type, and a fluctuation type. Experiments conducted on a real data set across different academic years and courses show that the proposed approach has better clustering accuracy and practicability. It is expected that this work may be used to strengthen the personalized learning support services system in educational institutions and develop a tool that integrates exploration and analysis work onto the web platform.