docker-mahout

Build mahout from source in a docker container

USAGE

After logging into docker:

docker run -it -v ``:/data/corpus:ro borromeotlhs/docker-mahout /bin/bash

assuming you've done the above, and that your corpus is segmented under self-labeled directories, you can train a Complementary NaiveBayes classifier on your corpus with:

$ /usr/local/mahout/bin/mahout seqdirectory -i /data/corpus -o /data/corpus-seq -xm sequential -ow

$ /usr/local/mahout/bin/mahout seq2sparse -i /data/corpus-seq -o /data/corpus-vectors -lnorm -nv -wt tfidf -ng 3 -n 2 --maxDFPercent 85

$ /usr/local/mahout/bin/mahout split -i /data/corpus-vectors/tfidf-vectors --trainingOutput /data/corpus-train-vectors --testOutput /data/corpus-test-vectors
--randomSelectionPct 40 --overwrite --sequenceFiles -xm sequential

$ /usr/local/mahout/bin/mahout trainnb -i /data/corpus-train-vectors -el
-o /data/model -li /data/labelindex -ow -c

(The above command line tells mahout, via the '-el' option, to extract labels and to store them, via the '-li' option, to ${WORK_DIR}/labelindex. You could, alternatively, utilize the '-l' option to provide your own csv file of labels to utilize on the input)

and will allow us to test with:

$ /usr/local/mahout/bin/mahout testnb -i /data/corpus-test-vectors -m /data/model -l /data/labelindex -ow -o /data/corpus-testing -c

TODO

You tell me ;)