Our team had initially planned to do an AWS machine learning configuration, but unfortunately our ML expert had to drop out. I have done a number of different information visualization projects, namely using d3.js, and have always found colorful, relatable, interactive displays to be the most meaningful. So rather than go the machine learning route, we went with a data-aggregation and visualization project.
What it does
This is a beginning of a dashboard application that can be used to visualize relationships between disparate sets of data that normally might not be compared with one another. This application becomes more powerful with more sets of data, but for the purpose of the hackathon, we were only able to include 5 data sets: 311 information, traffic signal locations, traffic incident locations, felony arrest locations, and domicile locations of registered sex offenders. The last 2 sets were included as an effort to address the worsening problems with human trafficking in Charlotte.
Challenges I ran into
The arrest records were quite incomplete. Most of the addresses were merely street names, and the sheer volume and variety of the data made visualization difficult. We pruned the data down to the more significant felony records, and incrementally fed the data through a web service to get Latitude and Longitude locations so they could be plotted on the map. This took about half of the time for one team member.