This project was the culmination of my 12-week data science bootcamp with Metis. For my project, I wanted to combine my love of fantasy football, social media and natural language processing. Can incorporating social media improve fantasy football projects? Can you build a sentiment tracker that allows fans and fantasy football players to see and compare the public's opinion on various players?
Twitter providers a fantastic API for access historical and streaming info. But the rest, or historical API only goes back a certain number of tweets, or a limited number of days. For this analysis, I needed to access multiple years of offseason tweets for model training, cross-validation and testing. I used my twitter scraping code build using the Selenium Python package to access the tweets about a list of player names over a specified period using Twitter's advanced search.
I obtaioned historical player stats from Armchair Analysis and fantasy football draft information from the MyFantasyLeague.com API.
For more information on the details and results, you can see a PDF copy of my presentation, or the keynote format with an embedded video of my player sentiment tracking tool.