Model Training

Once the training data is ready following the Sampling and creation of the dataset, the classification model can be trained.

We use a GBM based model. The default parameters will work well for the sample demo, but they can be changed inside the Model.java class

    gbm.ntrees = 1000;
    gbm.balance_classes = false;
    gbm.learn_rate = 0.1f;
    gbm.min_rows = 10;
    gbm.nbins = 20;
    gbm.cols =  new int[] {1,2,3,4,5,6};
    gbm.validation = ftest;

Make sure to rebuild the war file if you change model parameters as described here

Run the web application as described here

Use the training endpoint to trigger model training

curl http://thoth-predictor:port?action=trainModel

Model training status and eventual performance metrics can checked using the h2o web url at any point

http://thoth-predictor:54321

---

Step 1 : Sampling and Creation of the Dataset

Step 2 : Model Training

Step 3 : Exposing Trained Model via the API

Step 4 : Monitoring the Health of the Model

Step 5 : Predicting with the Model