Input/Output mapping sample

The model uses the following output fields:

• Age the age of the patient as integer
• BP the blood pressure as string (LOW,NORMAL,HIGH)
• Cholesterol the blood chlesterol value as string (LOW,NORMAL,HIGH)
• Na_to_K the natrium to Kalium ratio as float value
• Sex the patients sex as string (F,M)

The input stream schema for the model would look like this:

Predictors =
	int32 age,
	rstring sex,
	rstring bloodPressure,
	rstring cholesterol,
	float64 NA_to_K;

The model uses no output fields but one target named Drug. Without output fields defined, per default the model output contains the following feature values:

• PredictedValue the predicted value as string. This is the name of the drug to apply
• Confidence the confidence for the predicted value as float number
• Probability the probabilities for all the other possible output drugs as map with the drug names as keys and the confidence as values

The output stream schema for the model would look like this:

ScoringResult =
	rstring predictedValue,
	float64 confidence,
	map<rstring,float64> probabilities;

The complete operator invocation looks like this:

stream<Predictors,ScoringResult> DrugTreatments = PMMLScoring(InputRecords)
{
	param
		modelPath: getApplicationDir() + "/etc/drugNN_pmml_v1.xml";
		modelInputAttributeMapping: "Age=age,BP=bloodPressure,Cholesterol=cholesterol,Na_to_K=NA_to_K,Sex=sex";
		modelOutputAttributeMapping: "predictedValue=Drug.PredictedValue,confidence=Drug.Confidence,probabilities=Drug.Probability";
}