Post by Timothée Maurin

Finally done with uni. 3 years, a BSc in Data Science from UCL, capped off with a 79.10% (First Class) on my dissertation. Quick example of the idea: think about Uber fares. Most rides cost what you'd expect, but every so often a surge-priced airport run pays way more, meanwhile a fare can't really go below zero. The surprises are lopsided. Most models ignore this and build symmetric error bars anyway, wasting accuracy on a direction that barely happens. My dissertation (thanks to my supervisor, Dr F. Javier Rubio) builds prediction intervals that are lopsided on purpose, tight on the boring side, wider where the real uncertainty is. Worked better than expected, so I turned it into a Python package and a live demo too. Next up: BlackRock this August. For now, just grateful for the supervisors, mates, and one too many late nights in the library that got me here. 📄 Dissertation attached below, give it a read if you have time! Live demo: https://lnkd.in/e6KdFEmM #DataScience #UCL #ConformalPrediction #MachineLearning