This is a simple Bayesean application that uses your browsing history to estimate your gender.
The formula may or may not look familiar in this form, but it's simply Bayes theorem, as we all learned(?) in statistics.
It's interesting to look at the errors, and see what sites generated them. People seemed more likely to post their site list if they were misidentified.
Women evidently do more banking online.
chase.com 0.87
wamu.com 0.85
discovercard.com 0.87
Women are heavier users of search engines, and ask for directions more.
google.com 0.98
aol.com 0.82
ask.com 0.77
mapquest.com 0.83
Men are more likely to read newspapers online
nytimes.com 1.13
sun-sentinel.com 1.33
salon.com 1.07
Since I pay bills online and use multiple search engines, I end up rated at 66% female. But then my wife's history gets rated as slightly more male than female, so I guess we are compatible.
No comments:
Post a Comment