A bank typically gives rewards program to its card holders. People who sign up for the rewards program for some card are more likely to use that card. By looking at the past spending pattern of it's card users, a bank can tune it's rewards program so that it's card members are more likely to sign up for rewards and thus increase it's card usage.
What it does
We look at the spending pattern, where consumers go, and what products and business establishments are card members more likely to frequent.
How I built it
We downloaded JSON files using CapitalOne API. We used jupyter notebook to do the exploratory data analysis. We used folium package to integrate with OpenStreetMap.
Challenges I ran into
We were initially planning to use some machine learning techniques to cluster data, build customer/merchant networks, and make predictions about the data. However, ran into trouble downloading the JSON file from the start. We then tried to use MicroStrategy API to make some visualizations but that did not pan out. We had no time to invest on applying ideas from machine learning.
Accomplishments that I'm proud of
Successfully integrated consumer data with geo locations using leaflet API.
What I learned
Learn from data, explore data first before making assumptions.
What's next for Spending pattern of Capital One customers
Data provided by Capital One has a rich set of features that one can further explore. We did not touch upon time-series, ATM, branches, and many other topics that could add lot more value to what we want to model. One can use past data to find fraudulent/suspicious incidents, send alerts to users based on merchant location, etc.