Magazine article University Business

Driving Student Success with Predictive Analytics: Using Data to Identify At-Risk Students and Help Them Succeed

Magazine article University Business

Driving Student Success with Predictive Analytics: Using Data to Identify At-Risk Students and Help Them Succeed

Article excerpt

A University Business Web Seminar Digest * Originally presented on April 21, 2016

Why do students persist? Although there are some commonalities, the answer is different at every institution. Predictive modeling seeks to answer this question by discovering hidden relationships in data. By leveraging a clear picture of past and present behavior, predictive modeling uses statistical analysis to generate a confident simulation of future behavior. Higher education institutions can then use that insight to positively impact student trajectories and influence outcomes.

In this web seminar, an administrator from Crown College in Minnesota shared how the institution is using predictive analytics to identify at-risk students, and described the communication flow and intervention process used to leverage the information discovered from the predictive model. A student success expert from Jenzabar also highlighted real-world examples, best practices developed over many years of analyzing data, and some of the most popular risk factors and programs that have been developed to help students succeed.

Meghan Turjanica: At Jenzabar, when we're speaking of predictive analytics and predictive modeling, we are talking about creating custom predictive models. That means for every client that we work with, we look at their data specifically, and don't try to assign data to preexisting groups of categories. We don't assume that one algorithm is going to be predictive for every institution.

That's important to us. You have access to incredibly rich data, and some of it's very specific to your institution. And sometimes the results we get out of these models can be very surprising. For instance, at some institutions disciplinary factors end up being a positive for whether or not students stay or go. There are often data points like this that are specific to an institution, but that probably wouldn't be part of a model that was built off a general algorithm.

We are intentional and targeted when we are creating these models. We pay attention to cohort size. That means that we're looking for a group on your campus when we model for student success purposes. We're looking for a group that is large enough to have an impact but not so large that you won't have the resources to address the information that you gain. Our goal is to not just tell you who will be at risk, but to make sure that you have the resources to support those students and help them succeed.

When we're developing predictive models with you, we look at data. Institutions vary in understanding risk for their students. We have some institutions that have already started to build their own models. And we have other institutions that have only anecdotal stories about why students are at risk. So we work across a wide spectrum.

From there, once we know those predictive factors, we put them together and do a logistic regression that looks at the interplay of all factors so that the sum of all the factors ends up being much greater than the whole. So the individual parts themselves are predictive, but when we put them all together, you get an even stronger prediction.

[ILLUSTRATION OMITTED]

[ILLUSTRATION OMITTED]

From this logistical regression model, we find a retention score for each of the students. That information in and of itself is useful, but it's hard to take that information and understand what to do with it.

So we help you break those groups down into three threshold categories you will be able to work with: safe, at-risk and high-risk. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.