Advanced analytics and machine learning helped get to the bottom of this Ohio school’s concern: student visits to dean’s office.

The Business Need

It all started with a need. A need to cut back on the number of students visiting the dean’s office. A need to get to the bottom of an issue that spans schools and states.

At the heart of the matter: disciplinary issues. Because, the more time that teachers and school administrators spend on discipline, the less time they have to teach. It’s a problem that distracts from a school’s core function of educating students. It’s a problem that this middle school in Columbus, Ohio felt it was time to solve.

The high-performing charter school didn’t have the funds to hire a firm to investigate. Surely, nobody would do this work for free. Or, would they?

Enter Centric

A naturally curious, passionate team of our consultants heard the school’s story and felt compelled to help. So they volunteered, embarking on a mission to unlock new insights into this ongoing challenge using an innovative approach: advanced analytics and machine learning.

What technologies did they use? Microsoft’s Cloud Computing Platform, Azure as well as Azure Machine Learning Services, Excel, Power Pivot and SQL Server. The team applied their expertise in data and analytics to test various hypotheses.

They began by gathering data provided by the school. From anonymous student records to crime statistics for the school’s geographic area and even weather information. Thanks to Microsoft Azure, we were able to quickly run experiments on the data, which was key to maximizing learning.

Experiments included looking at: Weather patterns, temperatures, and humidity levels. Incidents of crime in nearby neighborhoods. Even patterns in disciplinary actions by students.

Experiments were implemented in two-week periods using agile software development techniques. With each iteration, the school generated new theories.

Then, the team applied techniques such as segmentation and linear regression to test that theory and present new findings. Each hypothesis ushered in a new wave of enthusiasm and learning.

The Results

Each hypothesis tested proved valuable – regardless of whether they identified a correlation or not. When a relationship in the data was shown, new insights were gathered.

When a relationship was not shown, myths about causes were dispelled. Together they learned that the data could not support neighborhood crime as a cause to a referral to the dean’s office. Neither could they find correlation with changes in the weather.

But some patterns emerged. One of the most impactful insights: a high percentage of repeat referrals typically occurred within one to three days of an initial referral.

To prevent future issues, the school moved to put a stronger focus on students who had recently had a referral.

In a first trimester update, referrals to the dean’s office at one of the school’s locations was down more than 40% overall, and down 73% for 6th graders. The school is optimistic that these findings coupled with future experiments will continue to reduce disciplinary visits to the dean’s office.