otter-stream-pmml

XML Standard

PMML (Predictive Model Markup Language) support via JPMML for portable model deployment.

Module Overview

The PMML module provides XML-based model format support via JPMML evaluator, enabling portable model deployment across different platforms. It supports various ML algorithms including regression, decision trees, neural networks, and ensemble models.

📋

PMML Engine

XML Standard

XML-based model format supporting various ML algorithms via JPMML evaluator.

  • Regression, decision trees, neural networks
  • Ensemble models and voting classifiers
  • Automatic data type conversion
  • Missing value handling
🔄

Data Preparation

Preprocessing

PMML's built-in data preparation and transformation capabilities.

  • Automatic field preparation via InputField
  • Value transformations from PMML definitions
  • Discretization and normalization
  • Outlier treatment

Implementing PMML Inference

Deploy portable PMML models across different platforms.

  1. Add Maven Dependencies 
    <dependency>
    <groupId>com.codedstreams</groupId>
    <artifactId>otter-stream-pmml</artifactId>
    <version>1.0.16</version>
</dependency>
<dependency>
    <groupId>org.jpmml</groupId>
    <artifactId>pmml-evaluator</artifactId>
    <version>1.6.3</version>
</dependency>
  2. Export Model to PMML 
    # From scikit-learn using sklearn2pmml
from sklearn2pmml import sklearn2pmml
from sklearn2pmml.pipeline import PMMLPipeline

pipeline = PMMLPipeline([...])
pipeline.fit(X_train, y_train)
sklearn2pmml(pipeline, "model.pmml")

# From R using pmml package
library(pmml)
model <- lm(formula, data)
saveXML(pmml(model), "model.pmml")
  3. Configure and Use Engine 
    ModelConfig config = ModelConfig.builder()
    .modelPath("model.pmml")
    .modelId("credit-scoring")
    .format(ModelFormat.PMML)
    .build();

PmmlInferenceEngine engine = new PmmlInferenceEngine();
engine.initialize(config);

Map<String, Object> inputs = Map.of(
    "age", 35,
    "income", 75000.0,
    "credit_score", 720
);

InferenceResult result = engine.infer(inputs);
double score = (double) result.getOutput("risk_score");

Quick Links

📋 PMML Engine Details 🔄 Data Preparation 📚 Full JavaDocs 🏠 Back to Home 🌳 ← XGBoost Module 🌐 Remote Module →

Maven Dependencies

<dependency>
    <groupId>com.codedstreams</groupId>
    <artifactId>otter-stream-pmml</artifactId>
    <version>1.0.16</version>
</dependency>
<dependency>
    <groupId>org.jpmml</groupId>
    <artifactId>pmml-evaluator</artifactId>
    <version>1.6.3</version>
</dependency>