A physicist in disguise: predicting loan defaults

Speaker: 
Fernando Nogueira
Event Date and Time: 
Thu, 2014-03-20 12:00 - 13:00
Location: 
Henn 318
Local Contact: 
Hal Clark
Intended Audience: 
Graduate

In this talk I will describe a remarkably fun project I recently embarked on - learning machine learning. This ongoing quest taught me many interesting things about a field of computer science called "machine learning", and ultimately led me to kaggle.com, where I just finished my first competition.

I am far from being an specialist, so I will try to focus on the tools I encountered along the way and the general picture I currently have in my mind (subject to change). I will describe in more detail a problem of how to predict the chance and amount of loan default based on historical banking data (noisy!), and how I went go about discovering these things.
 
Admittedly, if you are a theoretical physicist you probably will not need these types of tools in your research, while if you are an experimentalist, you may very well know a lot more than I do about the subject already. So my goal here is to give some perspective, how do you go from String Theory to predicting Walmart sales in one month, and why you should do it if you want to keep some doors open after you graduate.
Website development by Checkmark Media. Designed by Armada.

a place of mind, The University of British Columbia

Faculty of Science
Department of Physics and Astronomy
6224 Agricultural Road
Vancouver, BC V6T 1Z1
Tel 604.822.3853
Fax 604.822.5324

Emergency Procedures | Accessibility | Contact UBC | © Copyright The University of British Columbia