We have 2 buskers on the team (a busker is a street musician). While there are extremely successful street musicians, most are encounter difficulties due to 1) lack of skill or 2) tipping-averse audiences. While we can't do much for the former, spending propensity in any area is easily modeled by using transaction data. In order to help our fellow buskers, we built an app that points out the best areas to visit.

What it does

Buskr uses Capital One's Enterprise API to get information on transactions, including amount, date, and geographic location. Using a simple predictive ML model, Buskr creates a "spend propensity" index per person. Buskr then generates a user-friendly heat map of predicted tip revenue using these spend propensities and directs the user to the nearest "hot spot."

How we built it

We queried the data from Capital One's Enterprise API and parsed the jsons using Javascript. The data frames were then passed onto MongoDB, where they were linearly transformed into the final values used. We created the site using Node under a Bootstrap framework. We used Leaflet to create the maps and used a heat map extension for Leaflet to display the heat map.

Challenges we ran into

Capital One's randomized data had a few identification issues which presented issues when we tried to join tables.

What we learned

We learned about probability-based marketing models and how they apply to financial decisions. We also learned some interesting use cases for Javascript and how to use Leaflet with its extensions.

What's next for Buskr

Given more time, we would've introduced an online tipping feature for business to use to incentivize buskers to perform near their area.

Share this project: