[R] How to do knn regression?
xieyihui at gmail.com
Fri Sep 19 07:14:02 CEST 2008
I don't know any direct solutions to your question, but I don't think
it's difficult to write a few lines of code to find the k-nearest
neighbours for an observation with a missing value. Typically you need
the function dist() to compute distances, rank() or order() to find
the k-nearest neighbours, and finally using mean() or median() or any
statistic to make predictions.
To assure you the light work of programming, I can tell you all the
code of this example
(http://animation.yihui.name/dmml:k-nearest_neighbour_algorithm) is no
more than 100 lines :-D
But seriously speaking, I don't think my method is efficient. Maybe C
code will be much faster, as the knn() function in package 'class' has
Yihui Xie <xieyihui at gmail.com>
Phone: +86-(0)10-82509086 Fax: +86-(0)10-82509086
School of Statistics, Room 1037, Mingde Main Building,
Renmin University of China, Beijing, 100872, China
On Fri, Sep 19, 2008 at 10:17 AM, Shengqiao Li <shli at stat.wvu.edu> wrote:
> I want to do regression or missing value imputation by knn. I searched
> r-help mailing list. This question was asked in 2005. ksmooth and loess were
> recommended. But my case is different. I have many predictors (p>20) and I
> really want try knn with a given k. ksmooth and loess use band width to
> define neighborhood size. This contrasts to knn's variable band width via
> fixing a k. Are there any such functions I can use in R packages?
> Your help is highly appreciated.
> Shengqiao Li
> R-help at r-project.org mailing list
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
More information about the R-help