Merge f4a143d into a435235

lib/scoruby/model_factory.rb

@@ -4,6 +4,7 @@
 require 'scoruby/models/gradient_boosted_model/model'
 require 'scoruby/models/random_forest/model'
 require 'scoruby/models/naive_bayes/model'
+require 'scoruby/models/logistic_regression/model'
 
 module Scoruby
   class ModelFactory
@@ -17,10 +18,15 @@ def self.factory_for(xml)
       return Models::GradientBoostedModel::Model.new(xml) if gbm?(xml)
       return Models::DecisionTree.new(xml.child) if decision_tree?(xml)
       return Models::NaiveBayes::Model.new(xml) if naive_bayes?(xml)
+      return Models::LogisticRegression::Model.new(xml) if logistic_regression?(xml)
 
       raise MODEL_NOT_SUPPORTED_ERROR
     end
 
+    def self.logistic_regression?(xml)
+      !xml.xpath('PMML/GeneralRegressionModel').empty?
+    end
+
     def self.naive_bayes?(xml)
       !xml.xpath('PMML/NaiveBayesModel').empty?
     end

lib/scoruby/models/decision_tree.rb

@@ -17,7 +17,7 @@ def decide(features)
         while curr.children[0]
           prev = curr
           curr = step(curr, features)
-          return if didnt_step?(curr, prev)
+          break if didnt_step?(curr, prev)
         end
 
         curr.decision

lib/scoruby/models/logistic_regression/data.rb

@@ -0,0 +1,28 @@
+# frozen_string_literal: true
+
+module Scoruby
+  module Models
+    module LogisticRegression
+      class Data
+        COEFFICIENTS_VALUES_PATH = '//PMML/GeneralRegressionModel/ParamMatrix/PCell/@beta'
+        COEFFICIENTS_LABELS_PATH = '//PMML/GeneralRegressionModel/ParameterList/Parameter/@label'
+
+        def initialize(xml)
+          @xml = xml
+        end
+
+        def coefficient_values
+          @xml.xpath(COEFFICIENTS_VALUES_PATH).map{|attribute| attribute.value.to_f}
+        end
+
+        def coefficient_labels
+          @xml.xpath(COEFFICIENTS_LABELS_PATH).map(&:value)
+        end
+
+        def coefficients
+          @coefficients ||= Hash[coefficient_labels.zip(coefficient_values)]
+        end
+      end
+    end
+  end
+end

lib/scoruby/models/logistic_regression/model.rb

@@ -0,0 +1,32 @@
+# frozen_string_literal: true
+
+require 'scoruby/models/logistic_regression/data'
+require 'forwardable'
+
+module Scoruby
+  module Models
+    module LogisticRegression
+      class Model
+        extend Forwardable
+        def_delegators :@data, :coefficients, :coefficient_values
+
+        def initialize(xml)
+          @data = Data.new(xml)
+        end
+
+        def intercept
+          coefficient_values.first
+        end
+
+        def score(features)
+          logodds = intercept
+          features.each do |key, value|
+            logodds += coefficients[key.to_s] * value
+          end
+
+          1.0 / (1.0 + Math.exp(-logodds))
+        end
+      end
+    end
+  end
+end

spec/fixtures/logistic_regression.pmml

@@ -0,0 +1,43 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!--(Comment generated by ADAPA) PMML processed by ADAPA (Version : 4.3)-->
+<PMML version="4.3" xmlns="http://www.dmg.org/PMML-4_3" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.dmg.org/PMML-4_3 http://www.dmg.org/pmml/v4-3/pmml-4-3.xsd">
+ <Header copyright="Copyright (c) 2019 baileyjoseph" description="Generalized Linear Regression Model">
+  <Extension name="user" value="baileyjoseph" extender="SoftwareAG PMML Generator"/>
+  <Application name="SoftwareAG PMML Generator" version="1.4"/>
+  <Timestamp>2019-08-07 16:16:51</Timestamp>
+ </Header>
+ <DataDictionary numberOfFields="3">
+  <DataField name="value" optype="continuous" dataType="double"/>
+  <DataField name="prob" optype="continuous" dataType="double"/>
+  <DataField name="noise_var" optype="continuous" dataType="double"/>
+ </DataDictionary>
+ <GeneralRegressionModel modelName="General_Regression_Model" modelType="generalizedLinear" functionName="regression" algorithmName="glm" distribution="binomial" linkFunction="logit">
+  <MiningSchema>
+   <MiningField name="value" usageType="predicted" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="prob" usageType="active" invalidValueTreatment="returnInvalid"/>
+   <MiningField name="noise_var" usageType="active" invalidValueTreatment="returnInvalid"/>
+  </MiningSchema>
+  <Output>
+   <OutputField name="Predicted_value" feature="predictedValue" optype="continuous" dataType="double"/>
+  </Output>
+  <ParameterList>
+   <Parameter name="p0" label="(Intercept)"/>
+   <Parameter name="p1" label="prob"/>
+   <Parameter name="p2" label="noise_var"/>
+  </ParameterList>
+  <FactorList/>
+  <CovariateList>
+   <Predictor name="prob"/>
+   <Predictor name="noise_var"/>
+  </CovariateList>
+  <PPMatrix>
+   <PPCell value="1" predictorName="prob" parameterName="p1"/>
+   <PPCell value="1" predictorName="noise_var" parameterName="p2"/>
+  </PPMatrix>
+  <ParamMatrix>
+   <PCell parameterName="p0" df="1" beta="-2.88882584455418"/>
+   <PCell parameterName="p1" df="1" beta="6.08116002712647"/>
+   <PCell parameterName="p2" df="1" beta="-0.622996648302913"/>
+  </ParamMatrix>
+ </GeneralRegressionModel>
+</PMML>

spec/scoruby/model_factory_spec.rb

@@ -50,4 +50,13 @@
       expect(naive_bayes_model).to be_a Scoruby::Models::NaiveBayes::Model
     end
   end
+
+  context 'logistic_regression pmml' do
+    let(:logistic_regression_file) { 'spec/fixtures/logistic_regression.pmml' }
+    let(:logistic_regression_model) { Scoruby.load_model(logistic_regression_file) }
+
+    it 'loads LogisticRegression' do
+      expect(logistic_regression_model).to be_a Scoruby::Models::LogisticRegression::Model
+    end
+  end
 end

spec/scoruby/models/logistic_regression/data_spec.rb

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+
+describe Scoruby::Models::LogisticRegression::Data do
+  let(:xml) { Scoruby.xml_from_file_path(logistic_regression_file) }
+  let(:logistic_regression_file) { 'spec/fixtures/logistic_regression.pmml' }
+  let(:data) { described_class.new(xml) }
+
+  # it 'loads correct number of trees' do
+  #   expect(data.decision_trees.count).to eq 15
+  # end
+  #
+  it 'loads correct number of coefficient values' do
+    expect(data.coefficient_values.count).to eq 3
+  end
+
+  it 'loads correct number of coefficient labels' do
+    expect(data.coefficient_labels.count).to eq 3
+  end
+
+  it 'loads correct number of coefficients' do
+    puts data.coefficients
+    expect(data.coefficients.count).to eq 3
+  end
+end

spec/scoruby/models/logistic_regression/model_spec.rb

@@ -0,0 +1,21 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+
+describe Scoruby::Models::LogisticRegression::Model do
+  let(:logistic_regression_file) { 'spec/fixtures/logistic_regression.pmml' }
+  let(:xml) { Scoruby.xml_from_file_path(logistic_regression_file) }
+  let(:logistic_regression) { described_class.new(xml) }
+  let(:features) do
+    {
+      'prob' => 0.13,
+      'noise_var' => 0.5
+    }
+  end
+
+  context 'default' do
+    it 'scores features' do
+      expect(logistic_regression.score(features).round(8)).to eq 0.08243046
+    end
+  end
+end