Sorta obvious thing that was useful to me to say out loud: it's not usually useful to add features to a model that are highly correlated with existing features, unless they're also highly *uncorrelated* with those features in their availability.
-
-
-
Replying to @KentBeck @avibryant
Using a stylized example for bank underwriting: an entire raft of things correlated with FICO scores. Including them not generally good use of engineering time, except where they are more available than FICO. An uncorrelated thing: time spent on account application.
1 reply 0 retweets 1 like
(Time spent on account application has a very different distribution for consumers selected at random and people who have made a career out of applying for credit at banks, who are depending on the product either highly desirable or plausibly maximally undesirable.)
5:42 PM - 12 Feb 2019
0 replies
0 retweets
2 likes
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.