AUDIO | Predictive Analytics Support Success among At-Risk Student Populations (Part 1)
Facebook Twitter Linkedin Email Plusone Stumbleupon

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

Click here to download the first part of The EvoLLLution’s interview with Tim Renick.

AUDIO | Predictive Analytics Support Success among At-Risk Student Populations (Part 1)

By taking note of trends that have led to success and failure among past students, institutions have the capacity to develop a predictive analytics system that supports student retention, success and completion.


The following interview is with Tim Renick, vice provost and chief enrollment officer at Georgia State University. Renick spearheaded the adoption of a predictive analytics system at Georgia State aimed at improving retention and completion rates among at-risk, underserved student populations. In this first part of a two-part interview, Renick sheds light on the program and discusses the larger issue of retention among at-risk, non-traditional students.

Click here to read key takeaways.

1. What is the GPS (Graduate and Progression Success) program?

The GPS program is an attempt to deal with an issue we know has been critical to student success from the beginning of universities. If you go back to medieval times, the model that worked from the start was that of a mentor and mentee, an interaction between a teacher and student. We know from other avenues in post-secondary education — small liberal arts colleges with low ratios between students and faculty members and so forth — often these students excel, they get that personalized attention. The GPS advising system was an attempt to try to approximate that using technology on a much larger and more reasonable scale.

Georgia State is a big, messy, public university. We have 32,000 students, we’re right in downtown Atlanta and we have some really highly challenged students by most of the traditional indicators. About 60 percent of our students are on Pell Grants which means they come from household incomes of around $22,000 a year. Over 60 percent of our students are non-white, 40 percent of our students are first-generation college students. These are the students that American higher education has trouble succeeding with and, when you have large numbers of students who have a lot of need and you don’t have small student-to-faculty ratios, it seems like you might have a challenge with success.

What we’re trying to do with GPS advising is turn the odds in the students’ favor. What we have developed, using 10 years of Georgia State student data over 2.5 million grades, is predictive analytics for how our current students will do in individual classes and majors. These are not based on large categories like if they’re male or female or black or white or Hispanic, but instead based on their individual academic profile. What we’re really doing is mapping the current students’ profile against those of previous students over the past 10 years.

What does it take to succeed in a difficult upper-level chemistry course? What kind of grades do you need to achieve in the lower-level math courses or the prerequisite courses? These are very individualized questions we’ve never really had the capacity to answer at scale before. But now, with technology, we’re running these numbers on a daily basis.

We’re tracking current Georgia State students on a nightly basis based on 700 different alerts that can go off. What we’ve done is look at the 10 years of data to find out what kind of mistakes students have made in the past and how have they impacted the success of the students. Now we’re trying to give the current students the benefit of the wisdom of that experience.

If we have a student who struggled in a math course and they’ve signed up for an upper-level chemistry class, and we know by history that students with lower math skills are going to struggle in that chemistry class, we’re not sitting back and waiting for the student to take the chemistry class and get an F. The idea is now we’ll know immediately when the student signs up for that course that there is a risk factor and we will, within 48 hours, bring some kind of intervention in place. This is something we’ve never been able to do at scale before.

Modern technologies — the tracking systems we have, the predicative analytics, our ability to crunch some big data — have allowed us to do this in a way that is benefitting all our students and we’re seeing some really excellent results in the first few years of the system being in place.

2. Why did your team decide that predictive analytics were necessary for the success of your students?

Some of it is just common sense. Every student has an academic map; there are certain courses they need to take each semester to progress through that chemistry major. In the past, we would often find that even though the students have these maps in hand, they end up for various reasons signing up for other courses. They get confused and sign up for the wrong lab sequence or they may be swayed by the fact that a course is taught at a time that they don’t want to take the class or their friends are in another course and so forth.

Now with this system, an alert goes off during registration and if a student signs up for the wrong lab sequence, we immediately know about it and we intervene. We reach out to the student, we don’t force them to take any class, but we give them personalized information that we haven’t been able to in the past. In most cases the student moves their registration to the appropriate course. We did that more than 2,000 times last academic year alone. Two thousand students who were signed up for the wrong course got placed in the right course. This is a reassignment that’s occurring before the first day of class; we’re not shuffling them around so they’re falling behind during the first couple weeks of classes. We’re trying to streamline through some of these basic processes. When you’re dealing with 32,000 students, all making dozens of decisions every day — it was previously impossible to track them on an efficient basis. We would have had to have hundreds of advisors looking at all the actions of the students to be able to do the tracking we’re now doing in these simple ways.

We have so many students at Georgia State who are first-generation, low-income students and many of them have extremely worthy dreams but they don’t have a clear plan and a lot of practical examples, certainly, in their families for how to acquire those dreams, how to achieve them. What if instead of allowing students to fend for themselves, we began to reach out to them and really customize the kind of advice we’re giving them?

This interview has been edited for length.

- – - -

Key Takeaways

  • Georgia State University’s GPS System is an early-warning system based on predictive analytics that allows institutional administrators and counselors to intervene when a student is in a high-risk situation.
  • Rather than asking students to fend for themselves and  waiting for them to fail, administrators can get ahead of any issues and help put students on the path to academic success.

This is the first installment of a two-part interview with Tim Renick exploring Georgia State’s GPS system, its impact and the thought behind its development. In the second installment, Renick discusses how the system has created new staff efficiencies and improved retention and success metrics for the institution.



6 Requirements for Attracting & Retaining Non-Traditional Students in the 21st Century
 

Facebook Twitter Linkedin Email Plusone Stumbleupon
Opinions, Opportunities and Challenges, Retention

Tags:

Subscribe to The EvoLLLution, Get Premium Content And Stay Up To Date

4 Responses to AUDIO | Predictive Analytics Support Success among At-Risk Student Populations (Part 1)

  1. Helen G. Reply

    2014/08/01 at 12:24 pm

    Great piece on the potential of predictive analytics to assist students in making the right choices for their goals/needs. As more non-traditional students enter our system, some of whom have been out of school for decades, this type of personalized advice is a real value-added for them. Of course, analytics can also be helpful for historically underrepresented and first-generation students, as Renick notes. I’m interested to read the next piece on how this has impacted retention and completion at GSU.

  2. Fraser MacDonald Reply

    2014/08/01 at 3:17 pm

    Renick says factors such as gender, ethnicity and age aren’t used as indicators in the analysis. I’m curious to know what, then, is included in building individual academic profiles. He mentions course history and success in previous courses, but I wonder if he could list a few more.

  3. Miguel Gomez Reply

    2014/08/04 at 5:23 pm

    Fascinating piece on the potential for predictive analytics to assist both students and advisors in making more informed choices/giving better advice. My issue is that Georgia State was able to build on a decade of data culled from tens of thousands of students/graduates. What about smaller schools that don’t have that ready pool of data? How can we guarantee the validity of the data they collect? I think this comes down to whether institutions only benefit from predictive analytics if they can draw on their own experiences rather than more generalized ones. That’ll be an interesting piece to examine.

  4. Tim Renick Reply

    2014/08/15 at 12:00 pm

    Thanks for the feedback and questions. To establish our predictive analytics in academics for incoming freshmen, we are looking at things such a high school GPA, courses taken in prime academic areas and grades received, high school attended, and test scores. For the students once enrolled at Georgia State, we are looking at their college grades by individual courses, programs of studies, and majors. A “C” grade in math, for instance, might be fine for a humanities major, while it might represent a real warning sign for majors in STEM fields or Accounting. It is the data that determines: by running 10 years of Georgia State data, we are able to see via real historical evidence what kind of academic performance is predictive of overall success–and failure.

    For smaller colleges that may not have data sets large enough to produce accurate predictions, some states are compiling and sharing data among like institutions. If this is not happening in your state, national vendors such as Education Advisory Board–who we have worked with at Georgia State with good results–are compiling data from multiple institutions that may be of help. After all, many of these problems are not specific to one university: students with poor math skills are likely to struggle in science courses and students with poor writing skills are likely to struggle in English courses wherever they enroll.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

© 2014 The EvoLLLution. All rights reserved.