Kenny Shirley Statistician I’m a research scientist in the demand forecasting group at Amazon in New York City. My research interests (past and present) include hierarchical Bayesian modeling, MCMC methods, data and model visualization, text mining, and other topics related to applied statistics. This is my personal site, which is a mix of statistics research, side projects (mostly sports-related) and other stuff. |
|
NEWS 8/9/2017 I had a great time at JSM 2017 in Baltimore, MD, last week. I was the discussant in a session about applications of interactive graphics and I got to listen to three great talks by Ryan Hafen, Alicia Schep, and David Robinson. I reviewed some literature about interactive graphics, and I spoke a little bit about how we use interactive graphics on my team at Amazon to do demand forecasting. Here is a link to my slides. 6/26/2016: After 6.5 wonderful years in the statistics research department at AT&T Labs, I was ready for something new, and so I’ve joined the demand forecasting team at Amazon in New York City. I’ll be doing demand forecasting for retail products within the supply chain organization (aka SCOT) with colleagues that include Dean Foster and Lee Dicker. I’m excited for the new challenge! 1/12/2016: This is a very belated post to share the slides from my August 2015 JSM talk about my summarytrees R package. I've finally converted our code to compute and visualize summary trees (co-written with Howard Karloff) into an R package which contains a mixture of R, C, and javascript code. The GitHub repo for the package is here. It's not quite a finished product; I plan to clean up some of the internals before I submit it to CRAN. That being said, I'd love any comments, suggestions, or feedback on the current development version. For a couple of demos, check out one of these:
[Pictured: 18-node maximum entropy summary tree of DMOZ web directory] 5/14/2015: Next week Thursday (May 21st) my colleague Wei Wang will present our paper "Breaking Bad: Detecting Malicious Domains Using Word Segmentation" at the Web 2.0 Security and Privacy (W2SP) workshop in San Jose, CA. In the paper we describe how we segmented a set of domain names into individual tokens, and then used the resulting bag of words as an additional feature set to improve the predictive accuracy of our model to detect malicious domains. The outcome variable ("maliciousness") was gathered from the Web of Trust, a crowdsourced website reputation rating service. Some highlights of this project were:
For older news, click here. |