JPMML-SkLearn

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

Features

• Supported Estimator and Transformer types:
  • Clustering:
    • cluster.KMeans
    • cluster.MiniBatchKMeans
  • Matrix Decomposition:
    • decomposition.PCA
    • decomposition.IncrementalPCA
  • Discriminant Analysis:
    • discriminant_analysis.LinearDiscriminantAnalysis
  • Dummies:
    • dummy.DummyClassifier
    • dummy.DummyRegressor
  • Ensemble Methods:
    • ensemble.AdaBoostRegressor
    • ensemble.BaggingClassifier
    • ensemble.BaggingRegressor
    • ensemble.ExtraTreesClassifier
    • ensemble.ExtraTreesRegressor
    • ensemble.GradientBoostingClassifier
    • ensemble.GradientBoostingRegressor
    • ensemble.IsolationForest
    • ensemble.RandomForestClassifier
    • ensemble.RandomForestRegressor
    • ensemble.VotingClassifier
  • Feature Extraction:
    • feature_extraction.DictVectorizer
    • feature_extraction.text.CountVectorizer
    • feature_extraction.text.TfidfVectorizer
  • Feature Selection:
    • feature_selection.GenericUnivariateSelect (only via sklearn2pmml.SelectorProxy)
    • feature_selection.RFE (only via sklearn2pmml.SelectorProxy)
    • feature_selection.RFECV (only via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectFdr (only via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectFpr (only via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectFromModel (either directly or via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectFwe (only via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectKBest (either directly or via sklearn2pmml.SelectorProxy)
    • feature_selection.SelectPercentile (only via sklearn2pmml.SelectorProxy)
    • feature_selection.VarianceThreshold (only via sklearn2pmml.SelectorProxy)
  • Generalized Linear Models:
    • linear_model.ElasticNet
    • linear_model.ElasticNetCV
    • linear_model.Lasso
    • linear_model.LassoCV
    • linear_model.LinearRegression
    • linear_model.LogisticRegression
    • linear_model.LogisticRegressionCV
    • linear_model.Ridge
    • linear_model.RidgeCV
    • linear_model.RidgeClassifier
    • linear_model.RidgeClassifierCV
    • linear_model.SGDClassifier
    • linear_model.SGDRegressor
  • Naive Bayes:
    • naive_bayes.GaussianNB
  • Nearest Neighbors:
    • neighbors.KNeighborsClassifier
    • neighbors.KNeighborsRegressor
  • Pipelines:
    • pipeline.FeatureUnion
    • pipeline.Pipeline
  • Neural network models:
    • neural_network.MLPClassifier
    • neural_network.MLPRegressor
  • Preprocessing and Normalization:
    • preprocessing.Binarizer
    • preprocessing.FunctionTransformer
    • preprocessing.Imputer
    • preprocessing.LabelBinarizer
    • preprocessing.LabelEncoder
    • preprocessing.MaxAbsScaler
    • preprocessing.MinMaxScaler
    • preprocessing.OneHotEncoder
    • preprocessing.PolynomialFeatures
    • preprocessing.RobustScaler
    • preprocessing.StandardScaler
  • Support Vector Machines:
    • svm.OneClassSVM
    • svm.SVC
    • svm.NuSVC
    • svm.SVR
    • svm.LinearSVR
    • svm.NuSVR
  • Decision Trees:
    • tree.DecisionTreeClassifier
    • tree.DecisionTreeRegressor
    • tree.ExtraTreeClassifier
    • tree.ExtraTreeRegressor
• Supported third-party Estimator and Transformer types:
  • LightGBM:
    • lightgbm.LGBMClassifier
    • lightgbm.LGBMRegressor
  • SkLearn2PMML:
    • sklearn2pmml.EstimatorProxy
    • sklearn2pmml.PMMLPipeline
    • sklearn2pmml.SelectorProxy
    • sklearn2pmml.decoration.CategoricalDomain
    • sklearn2pmml.decoration.ContinuousDomain
    • sklearn2pmml.preprocessing.PMMLLabelBinarizer
    • sklearn2pmml.preprocessing.PMMLLabelEncoder
  • Sklearn-Pandas:
    • sklearn_pandas.CategoricalImputer
    • sklearn_pandas.DataFrameMapper
  • XGBoost:
    • xgboost.XGBClassifier
    • xgboost.XGBRegressor
• Production quality:
  • Complete test coverage.
  • Fully compliant with the JPMML-Evaluator library.

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.

Python installation can be validated as follows:

import sklearn, sklearn.externals.joblib, sklearn_pandas, sklearn2pmml

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

The JPMML-SkLearn side of operations

• Java 1.7 or newer.

Installation

Enter the project root directory and build using Apache Maven:

mvn clean install

The build produces an executable uber-JAR file target/converter-executable-1.3-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

Load data to a pandas.DataFrame object:

import pandas

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

First, instantiate a sklearn_pandas.DataFrameMapper object, which performs data column-wise feature engineering and selection work:

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

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

Second, instantiate any number of Transformer and Selector objects, which perform dataset-wise feature engineering and selection work:

from sklearn.decomposition import PCA
from sklearn.feature_selection import SelectKBest

iris_pca = PCA(n_components = 3)
iris_selector = SelectKBest(k = 2)

Third, instantiate an Estimator object:

from sklearn.tree import DecisionTreeClassifier

iris_classifier = DecisionTreeClassifier(min_samples_leaf = 5)

Combine the above objects into a sklearn2pmml.PMMLPipeline object, and run the experiment:

from sklearn2pmml import PMMLPipeline

iris_pipeline = PMMLPipeline([
    ("mapper", iris_mapper),
    ("pca", iris_pca),
    ("selector", iris_selector),
    ("estimator", iris_classifier)
])
iris_pipeline.fit(iris_df, iris_df["Species"])

Store the fitted sklearn2pmml.PMMLPipeline object in pickle data format:

from sklearn.externals import joblib

joblib.dump(iris_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 target/converter-executable-1.3-SNAPSHOT.jar --pkl-input pipeline.pkl.z --pmml-output pipeline.pmml

Getting help:

java -jar target/converter-executable-1.3-SNAPSHOT.jar --help

License

JPMML-SkLearn is licensed under the GNU Affero General Public License (AGPL) version 3.0. Other licenses are available on request.

Additional information

Please contact info@openscoring.io