feat: support classification with linear models

This corresponds to the LogisticRegression and RidgeClassifier classes from sklearn.
iamDecode · May 22, 2021 · 9650e25 · 9650e25
1 parent bb2fc93
commit 9650e25
Show file tree

Hide file tree

Showing 9 changed files with 475 additions and 103 deletions.
diff --git a/README.md b/README.md
@@ -20,19 +20,22 @@ $ pip install sklearn-pmml-model
 ## Status
 This library is in beta, and currently not all models are supported. The library currently does support the following models:
 
-| Model                  | Classification | Regression | Categorical features |
-|------------------------|----------------|------------|----------------------|
-| [Decision Trees](sklearn_pmml_model/tree) | ✅              |            | ✅<sup>1</sup>        |
-| [Random Forests](sklearn_pmml_model/ensemble) | ✅              |            | ✅<sup>1</sup>        |
-| [Gradient Boosting](sklearn_pmml_model/ensemble) | ✅              |            | ✅<sup>1</sup>        |
-| [Linear Regression](sklearn_pmml_model/linear_model) |                | ✅          |                      |
-| [Ridge](sklearn_pmml_model/linear_model) |                | ✅          |                      |
-| [Lasso](sklearn_pmml_model/linear_model) |                | ✅          |                      |
-| [ElasticNet](sklearn_pmml_model/linear_model) |                | ✅          |                      |
-| [Gaussian Naive Bayes](sklearn_pmml_model/naive_bayes) | ✅              |            |                      |
+| Model                                                  | Classification | Regression | Categorical features |
+|--------------------------------------------------------|----------------|------------|----------------------|
+| [Decision Trees](sklearn_pmml_model/tree)              | ✅             |            | ✅<sup>1</sup>        |
+| [Random Forests](sklearn_pmml_model/ensemble)          | ✅             |            | ✅<sup>1</sup>        |
+| [Gradient Boosting](sklearn_pmml_model/ensemble)       | ✅             |            | ✅<sup>1</sup>        |
+| [Linear Regression](sklearn_pmml_model/linear_model)   | ✅<sup>2</sup> | ✅         | ✅<sup>3</sup>        |
+| [Ridge](sklearn_pmml_model/linear_model)               | ✅             | ✅         | ✅<sup>3</sup>        |
+| [Lasso](sklearn_pmml_model/linear_model)               | ✅<sup>2</sup> | ✅         | ✅<sup>3</sup>        |
+| [ElasticNet](sklearn_pmml_model/linear_model)          | ✅<sup>2</sup> | ✅         | ✅                    |
+| [Gaussian Naive Bayes](sklearn_pmml_model/naive_bayes) | ✅             |            |                      |
 
 <sup>1</sup> Categorical feature support using slightly modified internals, based on [scikit-learn#12866](https://github.com/scikit-learn/scikit-learn/pull/12866).
 
+<sup>2</sup> These models differ only in training characteristics, the resulting model is of the same form. Classification is supported using `PMMLLogisticRegression`.
+
+<sup>3</sup> By one-hot encoding categorical features automatically.
 
 ---
 

diff --git a/models/generate_pmml.R b/models/generate_pmml.R
@@ -64,6 +64,22 @@ Pima.tr2
 
 
 
+library(glmnet)
+test = read.csv('/Users/decode/Developer/sklearn-pmml-model/models/categorical-test.csv', header=TRUE, sep=",")
+
+clf = glm(type ~., data=test, family=binomial)
+pmml_clf = pmml(clf, predicted_field = "type")
+saveXML(pmml_clf, "/Users/dennis/Downloads/naive_bayes.pmml")
+
+
+predict(clf, model.matrix(type ~ ., test)[,-1], type="raw")
+
+preds <- predict(clf, newdata = test)
+conf_matrix <- table(preds, test$type)
+conf_matrix
+
+
+
 
 library(gbm)
 library(r2pmml)

diff --git a/models/linear-regression.pmml → models/linear-model-lm.pmml b/models/linear-regression.pmml → models/linear-model-lm.pmml
diff --git a/models/linear-model-lmc.pmml b/models/linear-model-lmc.pmml
@@ -0,0 +1,92 @@
+<?xml version="1.0"?>
+<PMML version="4.4.1" xmlns="http://www.dmg.org/PMML-4_4" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.dmg.org/PMML-4_4 http://www.dmg.org/pmml/v4-4/pmml-4-4.xsd">
+ <Header copyright="Copyright (c) 2021 decode" description="Generalized Linear Regression Model">
+  <Extension name="user" value="decode" extender="SoftwareAG PMML Generator"/>
+  <Application name="SoftwareAG PMML Generator" version="2.4.0"/>
+  <Timestamp>2021-05-21 21:56:51</Timestamp>
+ </Header>
+ <DataDictionary numberOfFields="8">
+  <DataField name="type" optype="categorical" dataType="string">
+   <Value value="No"/>
+   <Value value="Yes"/>
+  </DataField>
+  <DataField name="npreg" optype="continuous" dataType="double"/>
+  <DataField name="glu" optype="continuous" dataType="double"/>
+  <DataField name="bp" optype="continuous" dataType="double"/>
+  <DataField name="skin" optype="continuous" dataType="double"/>
+  <DataField name="bmi" optype="continuous" dataType="double"/>
+  <DataField name="ped" optype="continuous" dataType="double"/>
+  <DataField name="age" optype="categorical" dataType="string">
+   <Value value="(20,30]"/>
+   <Value value="(30,40]"/>
+   <Value value="(40,50]"/>
+   <Value value="(50,60]"/>
+   <Value value="(60,70]"/>
+  </DataField>
+ </DataDictionary>
+ <GeneralRegressionModel modelName="General_Regression_Model" modelType="generalizedLinear" functionName="classification" algorithmName="glm" distribution="binomial" linkFunction="logit">
+  <MiningSchema>
+   <MiningField name="type" usageType="predicted" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="npreg" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="glu" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="bp" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="skin" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="bmi" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="ped" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="age" usageType="active" invalidValueTreatment="returnInvalid"/>
+  </MiningSchema>
+  <Output>
+   <OutputField name="Probability_Yes" targetField="type" feature="probability" value="Yes" optype="continuous" dataType="double"/>
+   <OutputField name="Predicted_type" feature="predictedValue" optype="categorical" dataType="string"/>
+  </Output>
+  <ParameterList>
+   <Parameter name="p0" label="(Intercept)"/>
+   <Parameter name="p1" label="npreg"/>
+   <Parameter name="p2" label="glu"/>
+   <Parameter name="p3" label="bp"/>
+   <Parameter name="p4" label="skin"/>
+   <Parameter name="p5" label="bmi"/>
+   <Parameter name="p6" label="ped"/>
+   <Parameter name="p7" label="age(30,40]"/>
+   <Parameter name="p8" label="age(40,50]"/>
+   <Parameter name="p9" label="age(50,60]"/>
+   <Parameter name="p10" label="age(60,70]"/>
+  </ParameterList>
+  <FactorList>
+   <Predictor name="age"/>
+  </FactorList>
+  <CovariateList>
+   <Predictor name="npreg"/>
+   <Predictor name="glu"/>
+   <Predictor name="bp"/>
+   <Predictor name="skin"/>
+   <Predictor name="bmi"/>
+   <Predictor name="ped"/>
+  </CovariateList>
+  <PPMatrix>
+   <PPCell value="1" predictorName="npreg" parameterName="p1"/>
+   <PPCell value="1" predictorName="glu" parameterName="p2"/>
+   <PPCell value="1" predictorName="bp" parameterName="p3"/>
+   <PPCell value="1" predictorName="skin" parameterName="p4"/>
+   <PPCell value="1" predictorName="bmi" parameterName="p5"/>
+   <PPCell value="1" predictorName="ped" parameterName="p6"/>
+   <PPCell value="(30,40]" predictorName="age" parameterName="p7"/>
+   <PPCell value="(40,50]" predictorName="age" parameterName="p8"/>
+   <PPCell value="(50,60]" predictorName="age" parameterName="p9"/>
+   <PPCell value="(60,70]" predictorName="age" parameterName="p10"/>
+  </PPMatrix>
+  <ParamMatrix>
+   <PCell targetCategory="Yes" parameterName="p0" df="1" beta="-57.1799981494652"/>
+   <PCell targetCategory="Yes" parameterName="p1" df="1" beta="0.722654058424025"/>
+   <PCell targetCategory="Yes" parameterName="p2" df="1" beta="0.170651218810002"/>
+   <PCell targetCategory="Yes" parameterName="p3" df="1" beta="0.455725762363011"/>
+   <PCell targetCategory="Yes" parameterName="p4" df="1" beta="-0.473218748281948"/>
+   <PCell targetCategory="Yes" parameterName="p5" df="1" beta="0.275493428386101"/>
+   <PCell targetCategory="Yes" parameterName="p6" df="1" beta="7.40623923752118"/>
+   <PCell targetCategory="Yes" parameterName="p7" df="1" beta="5.6829407356491"/>
+   <PCell targetCategory="Yes" parameterName="p8" df="1" beta="8.82062257424644"/>
+   <PCell targetCategory="Yes" parameterName="p9" df="1" beta="-4.44588099376691"/>
+   <PCell targetCategory="Yes" parameterName="p10" df="1" beta="-26.4273990722638"/>
+  </ParamMatrix>
+ </GeneralRegressionModel>
+</PMML>
diff --git a/sklearn_pmml_model/base.py b/sklearn_pmml_model/base.py
@@ -1,4 +1,5 @@
 from sklearn.base import BaseEstimator
+from sklearn.preprocessing import LabelBinarizer
 from xml.etree import cElementTree as eTree
 from cached_property import cached_property
 from sklearn_pmml_model.datatypes import Category
@@ -228,7 +229,11 @@ def __init__(self, pmml):
     PMMLBaseEstimator.__init__(self, pmml)
 
     target_type: Category = get_type(self.target_field)
-    self.classes_ = np.array(target_type.categories)
+    try:
+      self.classes_ = np.array(target_type.categories)
+    except AttributeError:
+      self._label_binarizer = LabelBinarizer(pos_label=1, neg_label=-1)
+      self._label_binarizer.classes_ = np.array(target_type.categories)
     self.n_classes_ = len(self.classes_)
     self.n_outputs_ = 1
 

diff --git a/sklearn_pmml_model/linear_model/__init__.py b/sklearn_pmml_model/linear_model/__init__.py
@@ -2,6 +2,14 @@
 The :mod:`sklearn_pmml_model.linear_model` module implements generalized linear models.
 """
 
-from .implementations import PMMLLinearRegression, PMMLRidge, PMMLLasso, PMMLElasticNet
+from .implementations import PMMLLinearRegression, PMMLLogisticRegression, PMMLRidge, \
+    PMMLRidgeClassifier, PMMLLasso, PMMLElasticNet
 
-__all__ = ["PMMLLinearRegression", "PMMLRidge", "PMMLLasso", "PMMLElasticNet"]
+__all__ = [
+    "PMMLLinearRegression",
+    "PMMLLogisticRegression",
+    "PMMLRidge",
+    "PMMLRidgeClassifier",
+    "PMMLLasso",
+    "PMMLElasticNet"
+]