How universities are using big data and predictive analytics to support students

Institutions are doing everything they can to increase the number of students enrolling in higher education.

However, at many institutions in the US, and at education providers in other regions of the world, only about half of the students who start university actually earn degrees.

To address the problem, many are turning to big data and predictive analytics to identify students in crisis and intervene before they walk out of the doors for the last time.

But while collecting data is important, putting data into practice seems to be the key to success.

The idea that low grades and poor attendance offer an early indicator for students dropping out is not new. But what is new is the ability to take large amounts of historical data, crunch it systematically and identify patterns that can be used methodically to identify students requiring additional support. And to put appropriate interventions in place to help students along the way.

The pros and cons of using predictive analytics

The benefits of bringing various datasets together to get a detailed view of each student's progress are clear.

The student can be supported more intelligently throughout their studies. Tutors can track their progress and be alerted to any individuals that are deemed at risk of early withdrawal from programs. And institutions benefit financially from more students completing their studies.

However, critics of this movement towards a more interventionist education point to the potential downsides of monitoring student data so closely. They worry about privacy, security, a centralized big-brother mentality, and an over-surveillanced education experience for students.

Additionally, no system is flawless, so some students requiring support do inevitably fall through the net.

And with more external organisations and consulting firms entering the market they might also question whether there are tight enough controls protecting the mountains of valuable student data being gathered.

Using data to identify students who might need help

At Georgia State University, in Atlanta, more students are graduating than ever before.

The institution is one of a growing number that has turned to data to help identify students who might be struggling, or predict when they might struggle with their student life and studies in the future, so they can provide support before students drop out.

Georgia State University's analytics system color codes a student's risk of dropping out. Sometimes the data is even used to predict potential course difficulties to start a dialogue between tutors and students to provide additional support or to influence a change to their selected programs.

The institution actually has a long track record with predictive analytics and interventions.

In the early days, when they initially reviewed their data to try to figure out what were the best advancing indicators that a student might drop out they thought they'd find a couple of dozen factors. Instead, they found 800!

A human might not uncover or resolve such complex patterns. But using data analytics the institution can now usefully use this information to identify students in need of support. And the institution has dramatically raised its graduation rates.

Other US-based universities, including the University of South Florida and Arizona State University, have also seen their graduation rates rise after turning to predictive analytics.

You can read more on their success story here: GSU Sucess 

Supporting students with intelligent interventions

Predictive analytics is only half the story, though.

The other key element to raise graduation rates are the interventions, the human element. Whether it be career advisors, or tutors or guidance counsellors reaching out, the proactive contact made with students at the right time based on relevant triggers is what seems to make a real difference.

And it's something that wasn't previously possible before predictive analytics reached the maturity levels we're experiencing today.

Georgia State University's analytics, for instance, combine a large number of factors to predict which students are needing human intervention. And they even identify students with strong academic records but who are late paying their tuition fees. As you'd expect, this has proven to be a strong indicator of financial difficulties. But rather than issuing a harsh penalty, the institution typically offers a last-minute tuition ‘grant' to cover a small outstanding bill.

Institutions can also bring together other data such as how many times students log on to campus systems or swipe into the library, for example, together with data around student performance to provide a very detailed picture. Some are even tracking how often students connect with the campus Wi-Fi and their behavior in terms of logging onto various cloud-based services.

By joining the dots across hard factors such as payment lags, examination outcomes, and attendance, with softer behavioral measures such as computer and Wi-Fi usage, the frequency of library visits, and the amount of time spent in certain areas on campus, institutions can ascertain how engaged in education and socially active students are. And they're armed to provide support like never before.

Personal data is becoming a key ingredient in the education experience

Whether you broadly support the use of predictive analytics to support students or see it as another case of algorithms taking over is up to you. But the use of data in this way looks likely to increase in the coming years as the pressure to recruit and retain students rises.

With a smaller pool of potential students, it won't be as easy for universities to replace each year's dropouts with new students. So schools will have a greater incentive to hold on to students.

Looking at this laterally, the future might not just be about students being better supported through human interventions based on predictive analytics.

The future might actually revolve around more personalized learning experiences, tailored to individual abilities and needs that could directly increase students' motivation and reduce their likelihood of dropping out in a more natural way.

We're still in the early stages of this phenomenon. But at TerminalFour we're taking a cautious but positive view on it, provided institutions are transparent with students about their approach and data handling techniques.

The outcome could be more personalized support and education experiences based on super-accurate data. It's an outcome that's bright and could amplify student success.

For more information you can read UCISA's report or have a look at this video from Educause