Among the various areas of the transaction-based Internet economy, online education – or “e-learning” – is perhaps most bountiful in its provision of minable, insight-bearing data. Every interaction between thousands of students and their institution is captured, and the resulting data can be aggregated, organized, and optimized to support predictive models that predict enduring popularity of courses of study, which disciplines demand the most resources from an online institution, and even student performance. In this case, Calexus Solutions helped one such online e-learning provider to predict with greater than ninety percent certainty when which students would drop out and why. This unlocked an opportunity for the online institution to decrease its attrition rate through highly-targeted communications.
Embedded AI-Based Analytics, Real-Time Scoring, and Translation of Findings
- Performed “leaky bucket” analysis as preliminary step to modeling attrition at various states of the online semester cycle;
- Attrition predictions based on over 7,000 variables including intra-semester test scores, assignment completion rates, demographic information, student-college interactions via the Internet, and weekday and time of online assignment submission;
- Adaptive modeling occurred continuously over multiple months so that attrition predictions could be continually refined;
- Using students’ previous academic scores and other qualifiable/quantifiable behavior, identified subpar student performance in mathematics as leading indicator of student attrition;
- Determined, among other conclusions, that one of every four students whose skill in mathematics was insufficient or below average would cease their studies altogether.