Profiling MOOCs from Viewing Perspective

Abstract

Course assessments inextricably connect to learning behaviors, which are different from traditional courses to Massive open on line courses (MOOCs). The emerged high course dropout and low exam participation make viewing be the mostly presentive behavior of MOOC learning, and make us call for new assessment indexes. Based on the viewing behavior data provided by course platform Icourse, we provide a method to recognize potential all-rounders, and indexes to measure course attractions and order correlations between teaching and learning. For example, we treat the video label of viewing events as a random variable, adopt its information to measure viewing width, and then use the geometric mean of the information and the relative length of viewing time comparing with the length of videos to measure course attractions. The index can measure the diminishing marginal utility for learners as well as the information increment by viewing new videos. However, the unclear relationship between viewing and learning makes our results could not be used to assess course quality positively. Nevertheless, these can be used to detect which aspects of a course need to improve.