Next: Four Spurious Effects?
Up: Evaluating Hypothesis and Experimental
Previous: Question 9
- 100% accuracy 1/1 or 100000/100000
- really want to know how predictive
- more stable statistic like
might be better - but can'
t use for cutoff (or Laplace accuracy or better yet a cost function) - definitely don't want for most datamining
- do we want for most learning?
Patricia Riddle
Fri May 15 13:00:36 NZST 1998