Exemplifying professionalism and a commitment to excellence, the Master of Science in Analytics (MScA) program’s summer 2021 Capstone Showcase featured an impressive field of ten teams, with two distinguished with Best in Show awards.
“All the hard work and perseverance was certainly on display in the exceptionally high quality of all the projects,” said Greg Green, executive director for the MScA program. “The projects spanned an array of industries, data types, and methodological approaches and I was especially pleased to see how inventive the teams were in developing solutions for their clients. All the projects count as impressive milestones and we are looking forwarded to staying connected with the graduates and hearing about their next achievements.”
As the culmination of the student experience in the MScA program, the Capstone Project unites the academic lessons learned in the classroom with a real-world problem. The students lead all aspects of the project with the support of faculty, staff, and subject matter experts. Summer 2021’s projects and presentations gained praise in particular for their leading-edge analytic solutions, effective storytelling, and use of data visualizations.
The members of the Best-in-Show teams and their projects were:
- Satya Aduri and Karlos Dodson for “Estimating the Claim Risk of Gallagher Basset Retail Stores”
- Martin Copello, Andrew Fliege, and Thomas Guardi for “First Floor Elevation Prediction Using CNN”
The projects spanned an array of industries, data types, and methodological approaches and I was especially pleased to see how inventive the teams were in developing solutions for their clients.Greg Green, Associate Senior Instructional Professor, Director MScA Program
Working with Gallagher Basset, a leader in innovation across risk and claims management, Aduri and Dodson, with their faculty advisor Don Patchell, focused on calculating risk for general liability claims. The challenge involved developing a model while using a bare minimum of data from the client. With access only to zip code and prior claims data, the first step to their solution involved combining 2018-19 claims data with publicly available data.
“We drew on client retail store data, census data, and weather data to build a model that identified risk segments,” the team wrote. “From there, we leveraged those segments in a second model to predict claims rate distribution in each segment. With our model, you can accurately predict risk for each store at the zip code level.”
With Ashish Pujari as their faculty advisor, Copello, Fliege, and Guardi’s project focused on the challenge of determining first floor height (FFH) when modeling risk for residential flood insurance. Traditionally gathered manually, a technique that tends to be time-consuming, costly, and inaccurate, the team developed a method to compute the FFH of residential properties using publicly available data.
“Using a Google API to gather street-view images of residential properties in flood-prone areas, we then labeled the images with bounding boxes and used them to train the YOLOv5 neural network,” the team wrote. “From there, we combined that model’s data with US Geological Survey data to generate accurate FFH calculations, which we incorporated into a Streamlit web app. We coupled that with a custom Python script to get batch predictions and analysis.”
In the end, all the Capstone presentations in the MScA’s summer 2021 Showcase received high praise from the judges and subject matter experts on hand. Of particular note was the creativity demonstrated by the graduates when it came to seeking out novel tools and technologies. They were also lauded for engaging with their data sets at the deepest levels possible.
Most important of all, of course, is the grit shown by the graduates in steering their projects to completion. In reaching such a milestone they could have delivered no better proof that they’re now ready to tackle whatever challenges the world outside the classroom throws their way.