Steven Levitt's legendary F r e a k o n o m i c s was the basis for my interest in exploring crime and other sociological and economic statistics. I wanted to apply machine learning and work with basic computational python libraries in a basic way for a generally useful application. The Facebook Messenger API also presents a powerful opportunity for mobile application since it already includes integrated access to data like location.
What it does
When a user enters an area of Philadelphia deemed 'statistically dangerous', an onboard Facebook Messenger app automatically sends a report logging location and time ready to be transmitted to authorities if the person becomes involved with a crime or can be verified as a witness to a crime taking place.
How I built it
Challenges I ran into
Understanding how the manipulate the dataset to achieve reasonable regression results was a main challenge in addition to setting up the NodeJS environment for a basic Facebook ChatBot.
Accomplishments that I'm proud of
I'm proud of being able to load and manipulate the dataset and troubleshoot through basic web app deployment.
What I learned
I learned that Python has great resources for scientific computation and a high level understanding of web applications.
What's next for Basic Crime and Machine Learning
Running more statistically robust results on the dataset can yield more interesting and realistic results. This dataset can possibly be cross-referenced with other datasets like housing to better understand urban inequality from a data-driven perspective. Next steps also include possibly implementing a way of managing Messenger such that the user can send location data automatically and efficiently.