Slow and Steady or Fast and Furious: An Analysis of Completion Duration in open.uom.lk

Abstract

The open.uom.lk platform is an open learning platform developed by the University of Moratuwa, Sri Lanka, offering free asynchronous online IT related courses. The platform was developed as an initiative to meet the growing human capital requirement of the IT industry in Sri Lanka, one of the only sectors showing strong growth even during the country’s recent economic crisis. In the two years since the release of its flagship course Trainee Full-Stack Developer, there have been over 270,000 enrolled learners. The course consists of six subjects covering Python programming, web design, and a professional practice module that includes a capstone project. Despite the massive number of enrollments, as in many other MOOCs, completion rates remain low. For instance, for the two beginner courses in programming and web design, into which all participants are enrolled automatically, completion rates stand at 7-10%. Completion rates for subsequent subjects are higher given that enrollment is conditional on completing prerequisites. From a planning perspective, it is not just estimates of how many participants complete the programs but also the time they take to do so that are integral for determining the future growth of the IT industry. As such, this study aims to develop a model for course completion using survival analysis which has the advantage of being able to model censored data (e.g. when duration is unknown due to non-completion) and provide forecasts of completion rates at different points in time. The analysis uses student activity completion reports and demographic information of participants including employment and educational background. The findings suggest that the course interaction variables, particularly the speed of completing early activities, are more important for predicting course completion than the demographic variables, though education and employment status are also significantly correlated with completion. The survival model could then be used for purposes of predicting both the completion probability as well as the timeframe within which the completion may be expected.