You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Stefano Bellasio <st...@gmail.com> on 2010/11/12 11:48:53 UTC

Use of seqdirectory and seq2sparse to SVD

Hi, im trying to convert my grouplens dataset "ratings.dat" with seqdirectory and then seq2sparse, but i had some problems:

1) seqdirectory and seq2sparse don't work if i don't add "-core" in my command line, without this Mahout tell me that is unable to load class (why??)
2) when i run seqdirectory it creates "chunk-0", it's right?
3) then i run seq2sparse and i obtain a dir called "vectors" (my output dir) with a serie of dir (tokenized-documents and so on).

My goal is to use SVD with Lanczos Solver to make recommendations with hadoop, and im looking for a tool to convert my dataset, this procedure is right? Now, when i try to call SVD from Mahout command line it say that i have no data under /vectors/tf-vectors/data 

Please give me a help ;) Thank you, regards