R help - Aug 2004 - help with knn from class library

Hi all,

I made some computations with the knn function from the class library.
If I execute this function several times (with the same parameters k, 
training set and test set), I obtain different results.
I don't understand why the results for my test set are different. Could 
you give me some explanations?
Is the solution for  a k-nearest-neighbor classifier  unique?

Best regards.

> From: Martin Olivier
> 
> Hi all,
> 
> I made some computations with the knn function from the class library.
> If I execute this function several times (with the same parameters k, 
> training set and test set), I obtain different results.
> I don't understand why the results for my test set are 
> different. Could 
> you give me some explanations?
> Is the solution for  a k-nearest-neighbor classifier  unique?
Could it be that you have ties in the data?  There's a `use.all'
argument:

 use.all: controls handling of ties. If true, all distances equal to
          the 'k'th largest are included. If false, a random selection
          of distances equal to the 'k'th is chosen to use exactly
'k'
          neighbours. 

So you may want to try that.  The other thing to try is to set.seed() before
calling knn().

Andy

 
> Best regards.