One of the things Cho asked for was precision rates.
As I was randomly selecting data points to test, I noticed that a lot of the entries were wrong. Looking more closely, I believe it may be an off-by-one error in addStartAndEndTimes. Looking at the code I don't see any obvious issue. It could be my batches of 500.
An approach is to shrink my dataset down to 500, then 600 or something, look at the results. At least at those sizes it'll be fast.
No comments:
Post a Comment