Contact Support Sign Up
Login  Sign up

The Illume Predictive Modeling Process

Civitas Learning’s dedicated team of data scientists create predictive models to deliver powerful insights about your institution’s student data. For Illume, the process starts with the extraction of your institution's historical data from your Student Information System, Learning Management System, and any other data sources you choose to include. Using that data, institution-specific models are created:

  1. Civitas Learning ingests data from your sources and creates additional data points, called derived variables.
  2. After historical data is collected, segmentation strategies are tested. Segmentation occurs to inform the creation of multiple models using different sets of variables depending on what may be predictive for different types of students. For example, the variables that are most predictive of whether an online, undergraduate student persists may be very different than the variables that are predictive for an on-ground, graduate student.
  3. The variables are chosen based on a hypothesis of what selection will deliver the strongest models for your institution. Quick testing reveals the predictive power of these models.
  4. If the initial hypothesis did not deliver strong results, new segmentation strategies are tested until the most effective models are identified.
  5. The final result is multiple institution-specific predictive models using the most predictive variables for different types of students. Each model is tested on a student data set to ensure accuracy and validity.

After testing, these models are ready for use with your active students. We ingest current student data from the same sources and apply our predictive models to deliver an individual persistence prediction for each student.

Surface Persistence Trends and Powerful Predictors

Persistence is the act of a student re-enrolling in at least one course for a future term at your institution and staying enrolled past the census date, typically 14 days after the start of term. After model creation, persistence information is determined:

  1. Your institution-specific predictive models generate a persistence prediction for every student. This prediction represents the likelihood that they will persist on to a future term selected by your institution.
  2. The average of every student’s persistence prediction is calculated. The average persistence prediction surfaces as the institution-wide persistence prediction, displayed in the top row as the persistence prediction for all students.
  3. As you add filters, the persistence prediction will change to reflect the same calculation for the students meeting the selected filter criteria, displayed in the bottom row as the persistence prediction for the active filter.

Finally, the top predictors of persistence are surfaced: 

  1. Variables with the most predictive power are determined. 
  2. The most predictive variables appear in Illume as Powerful Predictors. These will dynamically change as you add filters, as they will change depending on the students included.

Watch this video to learn more:

Learn how to use this data to direct outreach to students in: Feature Guide: Student Outreach

Did you find it helpful? Yes No

Send feedback
Sorry we couldn't be helpful. Help us improve this article with your feedback.