bitcamp

submissions etc project for Finra Challenge at Bitcamp Hackathon April 7-9, 2017

We used the Panda and spaCy api's for Python to go through court opinions, provided by Finra, and collect names of all those involved in cases and who they were involved with. We used spaCy's api specifically for entity identification, to tell which words parsed were actually names. We threw away any names containing special characters since the api would scrape raw HTML and sometimes mistake tags for entities. We used the networkx api to take that data and put it into a weighted graph. The graph weighted edges based on how often two people appeared together. Each node had a suspicion level attribute that increased based on if someone near them was convicted of some sort of crime, how closely related they were to that individual and if they were related to many people related to that individual. This model then would print to a file a list of names of people who pass a certain suspicion threshold that deemed them worthy of further investigation.