5 messages · Raji, Jannis, raji sankaran +1 more
Hi, I am using k means algorithm for clustering.My data contains a few null/NA values.kmeans doesnt cluster with those values.Are there any option like na.omit which can avoid these null values and cluster the remaining values? Thanks, Raji
View this message in context: http://r.789695.n4.nabble.com/Reg-null-values-in-kmeans-tp3085518p3085518.html Sent from the R help mailing list archive at Nabble.com.
I do not really understand your question. You can use use kmeans but without the observations that include the NA values (e.g. by deleting whole rows in your observation matrix). If you want to keep the information in the valid observations of those rows, I fear you need to look for a clustering algorithm that can handle missing values. I doubt that there is a kmeans version that can. Think about inserting means of all other observations into the gaps, though this introduces bias as well. Jannis Raji schrieb:
Hi, I am using k means algorithm for clustering.My data contains a few null/NA values.kmeans doesnt cluster with those values.Are there any option like na.omit which can avoid these null values and cluster the remaining values? Thanks, Raji
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20101216/b181ceba/attachment.pl>
Have your tried something like the following?
# put some data to cluster in a data.frame d <- data.frame(x1=log(1:50), x2=sqrt(1:50), x3=1/(1:50)) # put NA's in rows 1 and 3 d[1,1] <- d[3,3] <- NA # cluster the non-NA rows tmp <- kmeans(na.omit(d), 3) # 3 clusters # add cluster id vector to original dataset, aligned properly d$cluster <- rep(NA, nrow(d)) d[names(tmp$cluster), "cluster"] <- tmp$cluster head(d)
x1 x2 x3 cluster 1 NA 1.000000 1.0000000 NA 2 0.6931472 1.414214 0.5000000 3 3 1.0986123 1.732051 NA NA 4 1.3862944 2.000000 0.2500000 3 5 1.6094379 2.236068 0.2000000 3 6 1.7917595 2.449490 0.1666667 3
tail(d)
x1 x2 x3 cluster 45 3.806662 6.708204 0.02222222 1 46 3.828641 6.782330 0.02173913 1 47 3.850148 6.855655 0.02127660 1 48 3.871201 6.928203 0.02083333 1 49 3.891820 7.000000 0.02040816 1 50 3.912023 7.071068 0.02000000 1 Bill Dunlap Spotfire, TIBCO Software wdunlap tibco.com
-----Original Message----- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of raji sankaran Sent: Wednesday, December 15, 2010 7:43 PM To: Jannis Cc: r-help at r-project.org Subject: Re: [R] Reg : null values in kmeans Hi Jannis, Thank you for answering my question. I saw the option called na.omit when i used nnet() and tried to classify Iris data with that. I wanted to know if there is a similar option available in kmeans which can omit or in some way consider the null/NA values and cluster the observations.Currently, kmeans throws an error for the dataset with NULL/NA values.
From your answer, i could understand that, the option of
handling NULL/NA is not available with kmeans. Please correct me if am wrong. Thanks again :) On Wed, Dec 15, 2010 at 6:50 PM, Jannis <bt_jannis at yahoo.de> wrote:
I do not really understand your question. You can use use kmeans but without the observations that include the NA values (e.g.
by deleting whole
rows in your observation matrix). If you want to keep the
information in the
valid observations of those rows, I fear you need to look
for a clustering
algorithm that can handle missing values. I doubt that
there is a kmeans
version that can. Think about inserting means of all other
observations into
the gaps, though this introduces bias as well. Jannis Raji schrieb: Hi,
I am using k means algorithm for clustering.My data contains a few null/NA values.kmeans doesnt cluster with those values.Are there
any option like
na.omit which can avoid these null values and cluster the remaining values? Thanks, Raji
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Hi Raji, I am quite sure that kmeans in general is not able to handle missing values so most probably there wont be an option for this in R. Either you omit the observations with NAs as William proposed or you search for some algorithm that can handle missing values (not sure whether there is any). Other alternatives would be to put mean values in the NA places. This, however, biases the results. HTH Jannis raji sankaran schrieb:
Hi Jannis, Thank you for answering my question. I saw the option called na.omit when i used nnet() and tried to classify Iris data with that. I wanted to know if there is a similar option available in kmeans which can omit or in some way consider the null/NA values and cluster the observations.Currently, kmeans throws an error for the dataset with NULL/NA values.
From your answer, i could understand that, the option of handling NULL/NA is
not available with kmeans. Please correct me if am wrong. Thanks again :) On Wed, Dec 15, 2010 at 6:50 PM, Jannis <bt_jannis at yahoo.de> wrote:
I do not really understand your question. You can use use kmeans but
without the observations that include the NA values (e.g. by deleting whole
rows in your observation matrix). If you want to keep the information in the
valid observations of those rows, I fear you need to look for a clustering
algorithm that can handle missing values. I doubt that there is a kmeans
version that can. Think about inserting means of all other observations into
the gaps, though this introduces bias as well.
Jannis
Raji schrieb:
Hi,
I am using k means algorithm for clustering.My data contains a few
null/NA
values.kmeans doesnt cluster with those values.Are there any option like
na.omit which can avoid these null values and cluster the remaining
values?
Thanks,
Raji