Towards A General Method for Building Predictive Models of Learner Success using Educational Time Series Data

This paper presents a pedagogical and instructional-technology general method for building predictive models for education from time series log data. While it is common for models of learner achievement to include cognitive features, we instead are data mining only resource accesses in the learning environment. This has benefits in that the approach is inherently scalable to new contexts due to its data driven nature. While we have only just begun to apply these methods to our institutional Massive Open Online Course (MOOC) data, it shows promise as both a descriptive modeling technique as well as an engine for creating predictive early alerts.