Matching Moms to Mentors Through Machine Learning

A child’s early life experiences are early indicators of who they’ll become as adults. With parents being so influential during that time of development, a nonprofit wanting to help mothers partnered with University of Pittsburgh engineers to help expand their resources. 

NuturePA, a nonprofit organization which serves southwestern PA, centers its mission on connecting new mothers with trained mentors who offer support and guidance – all through text messaging.

“Our mentors provide valuable parenting information and activities for moms and babies, answer questions as they come up, offer guidance and advice, and provide a much needed shoulder to lean on during the critical first three years of a child's life,” explained Kate Brennan, NuturePA’s director of operations. “Nurture mentors build relationships with moms so that when needs arise, they are able to offer responsive support and, when necessary, provide referrals to our many community partners for additional services.”

With the goal of identifying participants’ potential level of desired support and matching them with appropriately trained mentors, the NurturePA team partnered with students at Pitt’s Swanson School of Engineering to determine a new, noninvasive tool to ensure that the program allocates resources effectively. 

Engineering Real Solutions for Real People 

When Gloria Givler, Tom Chimes, Nicole Lipa, and Andrew Polar signed up for the senior design class as part of their graduation requirement, they didn’t expect to work on a project that had such a real-world application. 

The senior design class is part of the SAINT Program through Pitt’s Department of Industrial Engineering. SAINT, or Sponsor an Industrial Engineering Team, is a vehicle through which senior industrial engineering student teams experience hands-on learning while working to find solutions to the real-life engineering and business problems of their sponsor organizations. 

“When I first learned about the partnership opportunity with NurturePA, I really wanted to be on this project,” said Givler. “Doing something important for society has always been important to me – particularly as an engineer.”

The team tested a number of classification algorithms based on data representing different demographic parameters. 

Each algorithm’s performance was evaluated across accuracy, recall (out of only high-need mothers, how many are correctly classified), and a custom “NPA Metric,” which combines the previous two while weighing recall more heavily. A logistic regression model performed the best and was incorporated into code scripts for prediction. 

When a new mother is enrolled in the program, their information can be sent to the model, and the predicted needs risk level is output to the NurturePA website. An updating script was also created. If more data is collected in the future, it can change the model based on that new data, aiming to improve performance. Additionally, a dashboard was constructed to provide visual insights into the model. It includes an interactive feature where users can select values for each parameter and the associated needs risk level prediction and probability of accuracy are displayed. 

As a result of the project, NurturePA can reach more mothers outside Allegheny County and ensure they receive the appropriate level of care and support.

Getting to the final product wasn’t without its challenges, though. The team cited their co-op program experiences in helping guide them through, but the majority of them weren’t familiar with Python, the programming language used for the machine learning model. 

“It was a great learning experience,” said Polar, one of the team members. “We really all got to pick up new skills that we wouldn’t have otherwise.” 

The other challenge was dealing with real data since it’s not easily accessible. 

“When you work with real-world data that affects real people, there are going to be bumps in the road,” Polar explained. “It’s very different from what we were used to in classroom settings where everything is mostly provided.” 

Even with the intensity of the subject material and work, the team had only minor conflicts along the way and ultimately utilized each other’s unique skill sets in designing the model. During their final presentation by NurturePA, the nonprofit organization was not only impressed with the tool the team designed, but their level of professionalism they demonstrated for such a young team. 

“This was our first experience working with a group of students and our team was blown away by what they were able to accomplish,” said Erica Cochran, director of development at NurturePA. 

Cochran continued that the nonprofit will be using the tool that Pitt students developed and that it has opened doors for the organization to extend their reach to more mothers in need – a goal that both the students and NurturePA aimed to achieve from the beginning. 

“I felt motivated from the very beginning knowing that we would help these individuals,” said team member Tom Chimes. “Now that I’ve graduated from Pitt, it’s rewarding to know that something I took part in developing is being used.”