Standarization before feature selection
2 views (last 30 days)
Show older comments
Hi Everyone, I have a general question: Is it necessary to standarize (e.g. zscore) the data before feature selection? And if yes - why?
Thank you in advance, Best regards Michael
1 Comment
mohan gopal
on 24 Aug 2011
ya feature extraction gives details ragarding the object area and background area as well it gives infm regarding the characteristics of the object , i hope its more beneficial to formulate a data base for feature selection in case of image processing
Accepted Answer
Arturo Moncada-Torres
on 24 Aug 2011
It all depends on your application. It is not a golden rule, but can be handy in some cases.
For example, let's suppose you will use these features for a machine learning application. If you are going to feed them to a C4.5 classification algorithm, I think it will not make a big difference because of the nature of the algorithm per se. However, if you are going to feed them to a Nearest-Neighbor classifier, it would be a good idea, since normalizing the data will allow the coordinated system to be more "congruent" (I can not find a better word).
So in a concrete sentence, it is not necessary, but it can be useful sometimes.
0 Comments
More Answers (0)
See Also
Categories
Find more on Get Started with Statistics and Machine Learning Toolbox in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!