Living Mutual by ‘pioneering’ a map

By Andrew Reagan, Tom Jeon, and Adam Fox
Members of the data science team at MassMutual.
Posted on Nov 28, 2017

Want to see something neat, especially if you live in or around Massachusetts?

Check this out (using Chrome). It’s an interactive data map of the Pioneer Valley, the center region of the state along the Connecticut River. Our data science team here at MassMutual created it for the Pioneer Valley Planning Commission (PVPC), a local agency responsible for increasing communication, cooperation, and coordination among government, business, and civic organizations in the area. 

The map project was the result of our Data Days for Good initiative, where we dedicate some of our team’s time and skillsets to fill in the gaps for MassMutual partner-organizations that advance social good.

You see, data science is more than a buzzword at MassMutual. To us, it means using science, engineering, technology, data, and partnership to enable data-driven decision-making. Sometimes that means utilizing data for big insurance projects or simply to avoid drinking bitter beer, but each part of the equation is important. In our group, we think about data science along three primary dimensions based on these disciplines, that are often too disparate for many organizations to leverage on their own.

So, through our Data Days for Good effort, the PVPC partnered with the MassMutual Data Science team to build a map-based tool for users to explore different regions of the Pioneer Valley. And, with hope, it will allow for some informed discussion and decision-making about various issues in those areas.

The project nuts and bolts

Our challenge was to effectively convey historical and geographical trends in demographics of the towns in the Pioneer Valley. To apply our skills and technology to the data that the PVPC collected, we used the same approach that our team has refined for longer-term projects at MassMutual.

For those interested in data science details, here’s the rundown.

Day 1: Data collection and analysis

We started with a raw CSV file (a “comma separated values” file, which allows data to be saved in a table), and set to work exploring how to work with the data itself. Since our goal was to build a web-based tool, we settled on using the Leaflet JavaScript library to display maps. We knew that we needed to parse the CSV data from a file into a JavaScript object that could be displayed on Leaflet.

Computer programmers have a habit of gravitating toward funny names, so we decided to use the “papaParse” package in JavaScript to read through the CSV data. While some team members focused on the data provided, others gathered the spatial data that would be needed to display the locality boundaries on a map.

Day 2: Modeling

One hallmark of projects that we take on in data science is that we build predictive models. This project doesn’t require us to make any predictions, but rather provides a tool for analysts to make decisions using data.

Just like how we would select variables for a model, we focused on how to build a tool that would allow exploration of these data variables. We decided to build the map on the left side of the web dashboard to show geographical trends, with other statistics on the right side to show temporal trends. By the end of the day, we built these visualizations and linked them together.

Day 3: Pilot

Our project teams are composed of members who have different roles, with data scientists concentrating on the statistical modeling and strategic consultants focusing on partnership, interpretation, delivery, and knowledge seeking.

Pilot projects are where delivery happens and the rubber hits the road; it’s where we put our work into action to assess the real-world impact of the predictive models that we built. For the pilot of the dashboard, we interacted with the tool ourselves and evaluated whether what we had built was able to provide us with a clear view of the dataset across many different dimensions.

After we were satisfied with the basic functionality, we sent the tool to the PVPC for feedback. And the PVPC liked the initial result.

This is just one example of the work we are doing to help our partner community groups. And as a team, we are excited at the prospect of pursuing other opportunities to engage our skillset with partners and within our communities. It’s part of our Live Mutual commitment, and we’re happy to live up to it.

More from MassMutual…

Data science at MassMutual: A grad’s view

School and work: MassMutual’s data science program