JPMML-SkLearn

Java library and command-line application for converting Scikit-Learn pipelines to PMML.

Table of Contents

• Features
  • Overview
  • Supported packages
• Prerequisites
  • The Python side of operations
  • The JPMML-SkLearn side of operations
• Installation
• Usage
  • The Python side of operations
  • The JPMML-SkLearn side of operations
• Documentation
• License
• Additional information

Features

Overview

• Functionality:
  • Three times more supported Python packages, transformers and estimators than all the competitors combined!
  • Thorough collection, analysis and encoding of feature information:
  • Names.
  • Data and operational types.
  • Valid, invalid and missing value spaces.
  • Descriptive statistics.
  • Pipeline extensions:
  • Pruning.
  • Decision engineering (prediction post-processing).
  • Model verification.
  • Conversion options.
• Extensibility:
  • Rich Java APIs for developing custom converters.
  • Automatic discovery and registration of custom converters based on META-INF/sklearn2pmml.properties resource files.
  • Direct interfacing with other JPMML conversion libraries such as JPMML-H2O, JPMML-LightGBM, JPMML-StatsModels and JPMML-XGBoost.
• Production quality:
  • Complete test coverage.
  • Fully compliant with the JPMML-Evaluator library.

Supported packages

Prerequisites

The Python side of operations

• Python 2.7, 3.4 or newer.
• scikit-learn 0.16.0 or newer.
• sklearn-pandas 0.0.10 or newer.
• sklearn2pmml 0.14.0 or newer.

Validating Python installation:

import joblib, sklearn, sklearn_pandas, sklearn2pmml

print(joblib.__version__)
print(sklearn.__version__)
print(sklearn_pandas.__version__)
print(sklearn2pmml.__version__)

The JPMML-SkLearn side of operations

• Java 1.8 or newer.

Installation

Enter the project root directory and build using Apache Maven:

mvn clean install

The build produces a library JAR file pmml-sklearn/target/pmml-sklearn-1.8-SNAPSHOT.jar, and an executable uber-JAR file pmml-sklearn-example/target/pmml-sklearn-example-executable-1.8-SNAPSHOT.jar.

Usage

A typical workflow can be summarized as follows:

1. Use Python to train a model.
2. Serialize the model in pickle data format to a file in a local filesystem.
3. Use the JPMML-SkLearn command-line converter application to turn the pickle file to a PMML file.

The Python side of operations

Loading data to a pandas.DataFrame object:

import pandas

df = pandas.read_csv("Iris.csv")

iris_X = df[df.columns.difference(["Species"])]
iris_y = df["Species"]

First, creating a sklearn_pandas.DataFrameMapper object, which performs column-oriented feature engineering and selection work:

from sklearn_pandas import DataFrameMapper
from sklearn.preprocessing import StandardScaler
from sklearn2pmml.decoration import ContinuousDomain

column_preprocessor = DataFrameMapper([
    (["Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width"], [ContinuousDomain(), StandardScaler()])
])

Second, creating Transformer and Selector objects, which perform table-oriented feature engineering and selection work:

from sklearn.decomposition import PCA
from sklearn.feature_selection import SelectKBest
from sklearn.pipeline import Pipeline
from sklearn2pmml import SelectorProxy

table_preprocessor = Pipeline([
    ("pca", PCA(n_components = 3)),
    ("selector", SelectorProxy(SelectKBest(k = 2)))
])

Please note that stateless Scikit-Learn selector objects need to be wrapped into an sklearn2pmml.SelectprProxy object.

Third, creating an Estimator object:

from sklearn.tree import DecisionTreeClassifier

classifier = DecisionTreeClassifier(min_samples_leaf = 5)

Combining the above objects into a sklearn2pmml.pipeline.PMMLPipeline object, and running the experiment:

from sklearn2pmml.pipeline import PMMLPipeline

pipeline = PMMLPipeline([
    ("columns", column_preprocessor),
    ("table", table_preprocessor),
    ("classifier", classifier)
])
pipeline.fit(iris_X, iris_y)

Recording feature importance information in a pickle data format-compatible manner:

classifier.pmml_feature_importances_ = classifier.feature_importances_

Embedding model verification data:

pipeline.verify(iris_X.sample(n = 15))

Storing the fitted PMMLPipeline object in pickle data format:

import joblib

joblib.dump(pipeline, "pipeline.pkl.z", compress = 9)

Please see the test script file main.py for more classification (binary and multi-class) and regression workflows.

The JPMML-SkLearn side of operations

Converting the pipeline pickle file pipeline.pkl.z to a PMML file pipeline.pmml:

java -jar pmml-sklearn-example/target/pmml-sklearn-example-executable-1.8-SNAPSHOT.jar --pkl-input pipeline.pkl.z --pmml-output pipeline.pmml

Getting help:

java -jar pmml-sklearn-example/target/pmml-sklearn-example-executable-1.8-SNAPSHOT.jar --help

Documentation

Integrations:

• Training Scikit-Learn GridSearchCV StatsModels pipelines
• Converting Scikit-Learn H2O.ai pipelines to PMML
• Converting customized Scikit-Learn estimators to PMML
• Training Scikit-Learn StatsModels pipelines
• Upgrading Scikit-Learn XGBoost pipelines
• Training Python-based XGBoost accelerated failure time models
• Converting Scikit-Learn PyCaret 3 pipelines to PMML
• Training Scikit-Learn H2O.ai pipelines
• One-hot encoding categorical features in Scikit-Learn XGBoost pipelines
• Training Scikit-Learn TF(-IDF) plus XGBoost pipelines
• Converting Scikit-Learn TF(-IDF) pipelines to PMML
• Converting Scikit-Learn Imbalanced-Learn pipelines to PMML
• Converting logistic regression models to PMML
• Stacking Scikit-Learn, LightGBM and XGBoost models
• Converting Scikit-Learn GridSearchCV pipelines to PMML
• Converting Scikit-Learn TPOT pipelines to PMML
• Converting Scikit-Learn LightGBM pipelines to PMML

Extensions:

• Extending Scikit-Learn with feature cross-references
• Extending Scikit-Learn with UDF expression transformer
• Extending Scikit-Learn with CHAID models
• Extending Scikit-Learn with prediction post-processing
• Extending Scikit-Learn with outlier detector transformer
• Extending Scikit-Learn with date and datetime features
• Extending Scikit-Learn with feature specifications
• Extending Scikit-Learn with GBDT+LR ensemble models
• Extending Scikit-Learn with business rules model

Miscellaneous:

• Upgrading Scikit-Learn decision tree models
• Measuring the memory consumption of Scikit-Learn models
• Benchmarking Scikit-Learn against JPMML-Evaluator
• Analyzing Scikit-Learn feature importances via PMML

Archived:

• Converting Scikit-Learn to PMML

License

JPMML-SkLearn is licensed under the terms and conditions of the GNU Affero General Public License, Version 3.0.

If you would like to use JPMML-SkLearn in a proprietary software project, then it is possible to enter into a licensing agreement which makes JPMML-SkLearn available under the terms and conditions of the BSD 3-Clause License instead.

Additional information

JPMML-SkLearn is developed and maintained by Openscoring Ltd, Estonia.

Interested in using Java PMML API software in your company? Please contact info@openscoring.io