> I would like to know how the classifier KNN deal missing values?
The following applies to the developer version of Weka...
KNN itself doesn't deal with missing values, it's the distance
function. E.g., in case of euclidean distance, the following applies
to missing values while computing the distance of the same attribute
between two weka.core.Instance objects:
- attributes in both Instance objects have missing value
=> 0
- only one value is missing
~ nominal attribute
=> 1
~ numeric attribute
=> 1 in case of normalization, maxRange-minRange for that
attribute otherwise
See method "difference" in the weka.core.NormalizableDistance class
(weka.core.EuclideanDistance is derived from this class).
Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/ Ph. +64 (7) 858-5174
_______________________________________________
Wekalist mailing list
Wekalist@...
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
