Re: How to use kmeans clustering algorithm of Mahout

[Paritosh Ranjan]

Can you explain something about the error and provide the stacktrace ?

On 12-09-2012 14:22, Don.Tan wrote:

The original data is here:

[hadoop@datamining ~]$ hadoop fs -ls /home/test/test
Found 1 items

-rw-r--r-- 1 hadoop supergroup 129213799 2012-09-12 15:45 /home/test/test/result

After I used "mahout seqdirectory -i /home/test/test/ -o /home/test/result/ -c UTF-8", get this:

[hadoop@datamining ~]$ hadoop fs -ls /home/test/result
Found 1 items

-rw-r--r-- 1 hadoop supergroup 129213898 2012-09-12 15:47 /home/test/result/chunk-0

And after "mahout seq2sparse -i /home/test/result -o /home/test/sparse":

[hadoop@datamining ~]$ hadoop fs -ls /home/test/sparse
Found 7 items

drwxr-xr-x - hadoop supergroup 0 2012-09-12 15:54 /home/test/sparse/df-count -rw-r--r-- 1 hadoop supergroup 442252 2012-09-12 15:53 /home/test/sparse/dictionary.file-0 -rw-r--r-- 1 hadoop supergroup 394853 2012-09-12 15:54 /home/test/sparse/frequency.file-0 drwxr-xr-x - hadoop supergroup 0 2012-09-12 15:53 /home/test/sparse/tf-vectors drwxr-xr-x - hadoop supergroup 0 2012-09-12 15:54 /home/test/sparse/tfidf-vectors drwxr-xr-x - hadoop supergroup 0 2012-09-12 15:53 /home/test/sparse/tokenized-documents drwxr-xr-x - hadoop supergroup 0 2012-09-12 15:53 /home/test/sparse/wordcount

Which should I do next? I used "mahout kmeans -i /home/test/sparse/ -o /home/test/kmeans -dm org.apache.mahout.common.distance.CosineDistanceMeasure -x 10 -k 20 -ow --clustering"

but I got error.....

Thx!


On 09/12/2012 03:24 PM, Paritosh Ranjan wrote:

I think you will need these two commands ( in the same order ) :

seqdirectory : Generate sequence files (of Text) from a directory
seq2sparse: Sparse Vector generation from Text sequence files

On 12-09-2012 12:28, Don Tan wrote:

I think I didn't explain clear enough and sorry for that.

The example showed before is a part of my data.

Each line is a user profile, for example, the first row is the features of

a user. And I want to apply k-means to this data.

I need to create a file saves all users profile as sparse vector and put

them in mahout k-means algorithm, how can I do that?

  Thanks for your advice!

Don Tan

2012/9/12 Paritosh Ranjan <pran...@xebia.com>

I could not understand the question correctly, can you explain more?
Here you can find how to use kmeans algorithm of Mahout

https://cwiki.apache.org/**confluence/display/MAHOUT/K-**Means+Clustering<https://cwiki.apache.org/confluence/display/MAHOUT/K-Means+Clustering>

.


On 12-09-2012 11:43, Don.Tan wrote:

Aloha!

I am new to hadoop and mahout, but I have set up the hadoop cluster.

I am working on a clustering task lately. I think I could not make it quickly because I don't know too much about how to deal with massive data ( my data contains 1400000 user and 50000 features..plus that is sparse ).

    Could you tell me how deal with that? A slice of data is here:

167555,152622,162252,79481,**66540,41942,75500,167898,**

61923,182083,180681,181135,**174449,166439,167307,174126,**87800,2826,

     98660,158620,33900,
4780,13922,45040,159210,26423,**1471,68200,70402,109721,**
145860,23740,5818,15087,47861,**158620,170482,170161,39120,**

164514,5854,169183,151229,**171110,163457,4356,21363,1307,**78105,1322,177011,167822,

176329,116300,175216,167307,**46710,138740,100681,2089,1842,**
1206,101702,99210,50460,89605,**177424,142901,176464,160625,**
38201,112101,4048,1716,167599,**140883,158250,175399,

     example above contains 4 user's data and each number is nominal
(denoting that is a kind of behavior of user, e.s, user 2 has
"98660","158620","33900" )

Please tell me how to work on that or which documents should I read..

     Thx!

    Don Tan