On: Command-line tools for data science

[aficionado]

A very powerful command line tool for Machine Learning: http://bigmler.readthedocs.org/en/latest/

Create a predictive model is as simple as: bigmler --train < data.csv

or create a predictive model for Bitcoin volume taking online data from Quandl and parsing it with jq in just one line.

curl --silent "http://www.quandl.com/api/v1/datasets/BITCOIN/BTCDEEUR.json" | jq -c ".data" | bigmler --train --field-attributes <(count=0; for i in `curl --silent "http://www.quandl.com/api/v1/datasets/BITCOIN/BTCDEEUR.json" | jq -c ".column_names[]"`; do echo "$count, $i"; count=$[$count+1]; done) --name bitcoin

More info here: http://blog.bigml.com/2013/01/31/fly-your-ml-cloud-like-a-ki...

[diziet → aficionado]

Hmm -- the model did not seem to spot something a simple linear regression did.

[aficionado → diziet]

In this particular case, I think that you are right. But isn't it powerful to have it at the command line level? It's great to quickly create models as you can also use them to make predictions at the command line level.

[jeroenjanssens → aficionado]

Impressive. I'll add it to the list.