Australian Schools Boost Student Success, Reduce Attrition by 13% – with AI

We see value in DataRobot AI Platform. We already see improvements in schools that have taken this to heart. Their curve is on the rise.

An Ambitious Vision: Data-Informed Student Success

With 80 schools and 44,500 students across New South Wales, Catholic Education Diocese of Parramatta (CEDP) holds a trove of data on its students, from performance to attendance to demographics. 

And in that data, Dr. Raju Varanasi, Director, Data Intelligence, saw an opportunity. What if CEDP could mine it to predict outcomes such as student achievement or attrition and take steps to improve it? AI, he believed, was the answer.

“We have 400 dashboards on every aspect of our business,” Dr. Varanasi said. “To unlock that data, we wanted automated machine learning because we don’t have the budget for data scientists.”

94% Prediction Accuracy

CEDP turned to DataRobot AI Platform to automate its predictive analytics from end to end. The platform stood out for its essential explainability, documentation, and the fact that, as a stand-alone platform, it would plug into CEDP’s existing data sources, mainly Alteryx.

CEDP began applying AutoML to its primary targeted use case: predicting achievement for students in their last year of high school. For that, CEDP could tap into more than a decade worth of data. 

DataRobot AI Platform simplifies the entire lifecycle, from model preparation to building to monitoring. With DataRobot’s Automated Feature Discovery, they weighed 40+ variables before settling on 26 that were truly informative, including past student performance, attendance, and demographics.

The platform then runs through numerous models, enabling Dr. Varanasi and a small team of analysts to determine which one is the most effective. Dr. Varanasi also appreciates the ability to drill down into individual student profiles to understand why models scored them a certain way.

Predicting – and Influencing – Student Success

Using the DataRobot AI Platform, CEDP found that it can predict student achievement at around 94 percent accuracy. This gives teachers and administrators confidence in using those insights to help at-risk students, or guide others to meet or exceed their expected performance, or aid them in choosing the subjects where they’ll excel. 

“We are doing something fundamental that every parent and student would like to,” Dr Varanasi said. “That is to answer, ‘With my track record so far in schooling, where will I end up if I choose history, math, chemistry, or biology?’”

CEDP wasn’t sure how educators would respond to the insights. But they surprised Dr. Varanasi by asking if his team could predict achievement two years out instead of one – for an even further head start on helping students. 

“Once we explained to teachers that we’re complementing their efforts, not substituting their efforts, the resistance barriers came down, the adoption grew, and the goodwill about data has grown,” Dr. Varanasi said. “And that’s a very fundamental shift.”

“Predictive analytics helps to unearth current trends in courses and predict future performance,” explained Robert Nastasi, Principal at Emmaus Catholic College. “This data can help us focus on preparing the whole student, no matter what pathway they wish to pursue.”

Reducing Attrition, Retaining Revenue

In applying AutoML to another challenge, student attrition, CEDP pulled in a variety of features thought to influence attrition, including attendance, missed payments, student engagement, and awards. A post-attrition survey for parents helped identify some of these key factors.

By predicting the students that might leave, CEDP takes action to encourage students to remain at CEDP schools, such as offering tuition assistance or other forms of support. With the help of these findings, the school reduced its attrition rate by about 13 percent. By retaining more than 100 students that would otherwise leave, the school also preserved around 1.2 million dollars.

There is a discernible change in school exits. The benefits can be converted into monetary terms, which more than pays for DataRobot AI Platform.

With success with these two use cases, Varanasi is now turning to the platform to understand why Wi-Fi strength and reliability vary across schools. Getting ahead of outages could avoid considerable lost productivity.

Making a Difference Every Day

CEDP is applying predictive analytics to foster a data culture. Even before bringing in the DataRobot AI Platform, CEDP had hundreds of data dashboards. Progress with student achievement and attrition spur teachers and staff to stay committed to developing a data-informed culture.

Beyond that, machine learning empowers Dr. Varanasi’s team despite not having data science backgrounds.

“DataRobot is like having 150 data scientists under my desk,” Dr. Varanasi says. 

The predictions made through the use of DataRobot AI Platform make a difference in keeping students engaged, successful, and in CEDP schools. 

“We see value in DataRobot AI Platform,” Dr. Varanasi said. “We already see improvements in schools that have taken this to heart. Their curve is on the rise.”