Current recommendation systems for stores work by logging online transactions, or by recording location timestamps in smart devices. We realized that these suggestions can be improved by utilizing credit card transaction histories to better improve the user experience.
What it does
By taking in a Capital-One bank account id, it uses the Nessie API to extract account transactions, merchants, and geolocation data. HoleInMyWallet can then give smart suggestions about what stores/merchants the user would like to shop at based on categories associated with each vendor as well as trends among users with similar shopping preferences. In addition, the app can display transaction histories on both a point-marker format and heatmap format for better data visualization.
How I built it
The app was build in Android Studio using the Google Maps api and backed by a firebase database for user data. The Nessie API is used to get information about transactions and vendors, and a Microsoft Azure cognitive-services cloud model is used to determine recommendation using machine-learning algorithms fed with usage data.
Challenges I ran into
Feeding properly formatted CSV files into the cognitive services model became an challenge, as there were several instances of failure when uploading improperly formatted data files to the Azure computing platform.
Accomplishments that I'm proud of
We successfully implemented an Android app and connected it to cloud computing platforms to create a novel user experience.
What I learned
We learned a lot about data analysis techniques as well as the intricacies of mapping large data sets.
What's next for HoleInMyWallet
HoleInMyWallet can expand to other aspects of transaction data analytics, such as analyzing weighted spending patterns to determine a more accurate financial model of the user. In addition, HoleInMyWallet may be able to offer descriptive, personalized monetary consulting for where and how to budget your money.