H2O AutoML ( [login to view URL] ) already trained.
We need to get a percent for prediction from command line.
Model is simple... input: 350+ integers&doubles, output is enum: 0 or 1.
Now it works on Google AutoML Tables, but we need to start work on H2O.
Command line request must be like:
java -jar [login to view URL] -model="[login to view URL]" -data="1,2,0.4,0.9,-3,............"
Result must be:
Or at JSON array, like: [Key:0, Probability: 0.3], [Key:1, Probability: 0.7]. Round is 4 digits after dot.
There a too much models (experiments). We need to make an automatization.
Input: path to model file, output: result in persent.
I cant give you a trained model for test and dev. You need to get an examples or thain something.
We use XGBoost and DeepLearning. But script mus work with all model types from latest stable H2O AutoML 188.8.131.52.
We need a first v.1.0 version of script. You may post them to opensource to your github account :)
But code must be structured and beautiful for continuin developing. With you (if you want), or without you (if you want to make a work onese).
H2O: [login to view URL]
Docs: [login to view URL]
POJO docs: [login to view URL]