Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint
Share this Page URL
Help

Chapter 7. Data Transformations > 7.7. Calibrating class probabilities

7.7 Calibrating class probabilities

Class probability estimation is obviously more difficult than classification. Given a way of generating class probabilities, classification error is minimized as long as the correct class is predicted with maximum probability. However, a method for accurate classification does not imply a method of generating accurate probability estimates: The estimates that yield the correct classification may be quite poor when assessed according to the quadratic (page 160) or informational (page 161) loss discussed in Section 5.6. Yet—as we have stressed several times—it is often more important to obtain accurate conditional class probabilities for a given instance than to simply place the instance into one of the classes. Cost-sensitive prediction based on the minimum expected cost approach is one example where accurate class probability estimates are very useful.


  

You are currently reading a PREVIEW of this book.

                                                                                                                    

Get instant access to over $1 million worth of books and videos.

  

Start a Free Trial


  
  • Safari Books Online
  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint