This project attempts to predict an opponent’s strategy in StarCraft using real, in-game observations through a Bayesian network model (I think that is a fair label, but let me know if you disagree!). The actual network is represented as a tree data structure where each node represents a step in...
[Read More]
Does YouTube Push Controversial Video Recommendations?
Natural Language Processing of YouTube Captions
YouTube has received criticism for the video recommendations it gives after a user watches a video on the site. Critics allege that YouTube has prioritized viewership and engagement over all else. As a result, recommended videos are often more controversial than a viewer’s initial search and likely to contain misleading...
[Read More]
Oil and Gas Data Challenges
Analyzing Completion Uplift in West Texas Wells
One aspect of my previous job was the financial modeling of development programs for new wells. Modern wells are typically hydraulically fractured (“fracked”) to improve well productivity. I wanted to evaluate the relationship between completion design and the expected performance uplift to well, taking into account variables such as proppant...
[Read More]
Which NYC Stations are the Most Busy?
New York Metro Commuter Analysis
I recently started at Metis, an intensive 12-week data science-focused bootcamp, at their Chicago campus. Our first project was to make sense of some MTA (New York City’s subway authority) commuter data on behalf of our “client” who was hoping to promote an upcoming women-in-tech focused gala at a select...
[Read More]