Predicting with the Model

A valid instance of a ModelTrainer.java class, Cloud.java and an instance of a Solr request are required for prediction purposes.
An example can be found in the PredictionDemo.java class

• Create a model trainer instance and specify the URL of the Thoth Predictor. Initializing the class will also boot the H2o cloud and load the pre trained model

  ModelTrainer modelTrainer = new ModelTrainer();
  modelTrainer.setThothPredictorUri(THOTH_PREDICTOR_URL);
  modelTrainer.init();
  modelTrainer.refresh();

• Create an instance from a Solr request using the helper functions

    String solrRequest = "q=*:*"
    double[] instance = ArrayUtils.toPrimitive(
      TruliaInstance.create(
        TruliaConverter.solrQueryToQueryPojo(solrRequest, null, null, mapper)
      , SLOW_QUERY_THRESHOLD, true)
    );

• Use the Cloud class to fetch the prediction probabilities

float prediction = Cloud.predict(instance);

• Use the threshold connected to the current model to determine the prediction label

boolean isSlowQuery = (prediction < modelTrainer.getThreshold() )? false : true;

---

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