Hello all,
I've searched through examples and so far have seen examples on how to do to
use one-hot-encoder only for model fitting or for evaluator, but can't
figure out how to do this for the predict call. For example, we see use of
one-hot as inputs to :
1. RF_MODEL = trainer.fit(
<ignite>,
<trainingcache>, // this has category column before
one-hot
split.getTrainFilter(),
<one-hot-encoder> // this does one-hot inside the model
- how do I get the cache with additional columns?
);
OR ALSO here:
2. RegressionMetricValues regMetrics = Evaluator.evaluateRegression(
<trainingcache>,
split.getTestFilter(),
<rf_model>
<one-hot-encoder>
);
But rfmodel.predict(Vector features) requires the original Vector with
categorical columns be already converted into all doubles. What is best way
to do this intermediate step.
--
Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
