Well there are dozens of tool available for that now. Just read some recent reviews:

http://bib.oxfordjournals.org/content/early/2012/11/22/bib.bbs075.full

http://www.nature.com/nrg/journal/v13/n4/full/nrg3162.html

http://www.mdpi.com/1422-0067/14/4/8179/pdf‎

http://journal.imbio.de/articles/pdf/jib-127.pdf‎

In general, my personal preference would be to use known associations rather than predicted associations: the lists of TargetScan targets can get pretty large.

If you have an institutional license, I think IPA provides the best tool for miRNA targets (both known and predicted). If you have mRNA data, it can also filter the list based upon mRNA and miRNA expression (which I think is really important, if you are working with predicted targets).

If you don't have an IPA license, I think TarBase is the next best thing (for known miRNA targets):

http://diana.imis.athena-innovation.gr/DianaTools/index.php?r=tarbase/index

Different miRNA target prediction programs produce different results and have high false positive rates. You can use CLIP-Seq datasets to further filter the predicted miRNA targets. starBase collect and curate 111 CLIP-Seq (PAR-CLIP, CLASH, HITS-CLIP, ICLIP) datasets.

You can download them at download page: http://starbase.sysu.edu.cn/download.php

You should consider these tools in the light of new data from the CLASH experiment. It turns out that most of the assumptions used for target predictions are not accurate.
A prediction tool with a different approach than the seed is RepTar. For some reason it was left aside in many discussions.