Machine learning and analytics to derive insights from student behaviors.
To understand which students are at-risk and which students are not, you cannot look at behaviors in a vacuum. Behaviors that drive success will be different for individual universities and student groups. To effectively analyze which students are on track, you have to compare them against the right peers, and understand what made those peers successful.
Analyze. Compare. Predict.
Contextualize Behaviors to Correctly Identify At-Risk Students
- Attendance Percentage 88% 88%
- Library Time 22% 22%
- On-Campus Time 83% 83%
- Social Activities 54% 54%
- LMS Activity 56% 56%
- Attendance Percentage 90% 90%
- Library Time 78% 78%
- On-Campus Time 65% 65%
- Social Activities 53% 53%
- LMS Activity 87% 87%
Predict Success With Higher Certainty
Determing appropriate peer groups is the paramount of measuring student success. Art students don’t behave like physics students, on-campus students don’t behave like off-campus students, and rural college students don’t behave like urban university students.
The first step in our machine learning process is to cluster students and create appropriate cohorts. Only then can we understand the importance of student behaviors.
Machine Learning on Historical Success
We do not presume any behaviors to be positive or negative in determining student success. Once a peer group is established, we train machine learning models with historical data to understand which behaviors are important to which cohorts.
Armed with this information, we can inform advisors which behaviors they should encourage and which students have developmental opportunities.