scoruby 0.1.3 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 98eb545b1bb149f22fb5afbaf467d144af1dbb2c
-  data.tar.gz: dbae0fa59acbf5bb10252bd1971cda46fafb8095
+  metadata.gz: f1a4d1fcce322b50113aa532d34ea7397e17dfa7
+  data.tar.gz: e39e4e849ac6c98c14625372e1e17219b7ae2999
 SHA512:
-  metadata.gz: c572f65fdd04226519f6518aa56ec92d942b8882c700828f2c17447b89cff9286fea243a8f171b2b259b3a15dbce6985ea474ff14a802894fa068546db355ce8
-  data.tar.gz: 9bd8096adc82636f5a10d6b93d6f0230a7a82b67a755da53153b32dc58c29887a06a31af86235bf2533792bdbccc8b9c77d9677bb3c099c8ab30793aa4696ce4
+  metadata.gz: bdd1e14f38a6ab1675c54e7b795bc56f7bab845f6b05f3f8725c1b439737e23471dec7a3f94b60d982f8dbdef349a892d4c045c2f48ff45e0b0c06057599e86e
+  data.tar.gz: ec3ae16a4b3fd4b6299f89c925718ad9583fca0d7d27dc45661edc5ea8597a183869a29552aeaa74ddb01af92f421144a89cf1adedf93527e0d0811855c3c0f0

data/README.md

@@ -1,10 +1,17 @@
 <a href="https://codeclimate.com/github/asafschers/scoruby"><img src="https://codeclimate.com/github/asafschers/scoruby/badges/gpa.svg" /></a>
-[![Gem Version](https://badge.fury.io/rb/random_forester.svg)](https://badge.fury.io/rb/random_forester)
+[![Coverage Status](https://coveralls.io/repos/github/asafschers/scoruby/badge.svg?branch=master)](https://coveralls.io/github/asafschers/scoruby?branch=master)
+[![Gem Version](https://badge.fury.io/rb/scoruby.svg)](https://badge.fury.io/rb/scoruby)
 [![Build Status](https://travis-ci.org/asafschers/scoruby.svg?branch=master)](https://travis-ci.org/asafschers/scoruby)
 
 # Scoruby
 
-Reads Random Forest PMML files and creates Ruby Random Forest classifier model.
+Ruby scoring API for Predictive Model Markup Language (PMML).
+
+Currently supports random forest and gradient boosted models.
+
+Will be happy to implement new models by demand, or assist with any other issue.
+
+Contact me here or at aschers@gmail.com.
 
 ## Installation
 
@@ -23,14 +30,115 @@ Or install it yourself as:
     $ gem install scoruby
 
 ## Usage
+### Random Forest
+#### Generate  PMML - R
+
+```R
+
+# Install and require randomForest, pmml packages
+
+install.packages('randomForest')
+install.packages('pmml')
+library('randomForest')
+library('pmml')
+
+# Login to Kaggle and download titanic dataset 
+# https://www.kaggle.com/c/titanic/data 
+# Load CSV to data frame -
+
+titanic.train <- read.table("titanic_train.csv", header = TRUE, sep = ",")
+titanic.train$Survived <- as.factor(titanic.train$Survived)
+
+# Train RF model
+
+titanic.rf <- randomForest(Survived ~ . - Name - Cabin - Ticket,
+                           data = titanic.train, 
+                           na.action = na.roughfix)
+
+# Generate pmml from model
+
+pmml <- pmml(titanic.rf)
+saveXML(pmml, 'titanic_rf.pmml')
+
+```
+
+#### Classify by PMML - Ruby
 
 ```ruby
-random_forest = Scourby.get_model 'rf.pmml'
-features = {a: 1, b: true, c: "YES"}
+
+random_forest = Scoruby.get_model 'titanic_rf.pmml'
+features =  {
+        Sex: 'male',
+        Parch: 0,
+        Age: 30,
+        Fare: 9.6875,
+        Pclass: 2,
+        SibSp: 0,
+        Embarked: 'Q'       
+    }
+
 random_forest.predict(features)
+
+=> "0"
+
 random_forest.decisions_count(features)
+
+=> {"0"=>441, "1"=>59}
+
 ```
 
+### Gradient Boosted model
+
+#### Generate  PMML - R
+
+```R
+
+# Install and require gbm, r2pmml
+
+library("devtools")
+install_github(repo = "jpmml/r2pmml")
+
+library("r2pmml")
+library("gbm")
+
+# Login to Kaggle and download titanic dataset 
+# https://www.kaggle.com/c/titanic/data 
+# Load CSV to data frame -
+
+titanic.train <- read.table("titanic_train.csv", header = TRUE, sep = ",")
+titanic.train$Survived <- as.factor(titanic.train$Survived)
+
+# Train GBM model
+
+titanic.gbm <- gbm(Survived ~ . - PassengerId - Name - Cabin - Ticket,  data = titanic.train)
+
+# Generate pmml from model
+
+pmml <- r2pmml(titanic.gbm, 'titanic_gbm.pmml')
+
+```
+
+#### Classify by PMML - Ruby
+
+```ruby
+
+gbm = Scoruby.get_model 'gbm.pmml'
+
+features =  {
+        Sex: 'male',
+        Parch: 0,
+        Age: 30,
+        Fare: 9.6875,
+        Pclass: 2,
+        SibSp: 0,
+        Embarked: 'Q'       
+    }
+
+gbm.score(features)
+
+=> 0.3652639329522468
+
+```
 
 ## Development
 

data/lib/random_forest/decision_tree.rb

@@ -23,14 +23,20 @@ class DecisionTree
   private
 
   def step(curr, features)
-    curr = curr.left if curr.left && curr.left.true?(features)
-    curr = curr.right if curr.right && curr.right.true?(features)
+    curr = step_on_true(curr, features, 0)
+    curr = step_on_true(curr, features, 1)
+    curr = step_on_true(curr, features, 2)
+    curr
+  end
+
+  def step_on_true(curr, features, num)
+    return curr.children[num] if curr.children && curr.children[num] && curr.children[num].true?(features)
     curr
   end
 
   def didnt_step?(curr, prev)
     return false if (prev.pred != curr.pred)
-    Scoruby.logger.error "Null tree: #{@id}, bad feature: #{curr.left.pred.field }"
+    Scoruby.logger.error "Null tree: #{@id}, bad feature: #{curr.children[0].pred.field }"
     true
   end
 end

data/lib/random_forest/features.rb

@@ -0,0 +1,16 @@
+class Features
+
+  attr_reader :formatted
+
+  def initialize(features)
+    @formatted = format_booleans(features)
+  end
+
+  def format_booleans(features)
+    features.map { |k, v|
+      features[k] = 'f' if v == false
+      features[k] = 't' if v == true
+    }
+    features
+  end
+end

data/lib/random_forest/gbm.rb

@@ -0,0 +1,29 @@
+require 'decision_tree'
+require 'features'
+
+class Gbm
+  GBM_FOREST_XPATH = '//Segmentation[@multipleModelMethod="sum"]/Segment'
+  CONST_XPATH      = '//Target/@rescaleConstant'
+
+  def initialize(xml)
+    @decision_trees = xml.xpath(GBM_FOREST_XPATH).collect{ |xml_tree|
+      DecisionTree.new(xml_tree)
+    }
+    @const = Float(xml.xpath(CONST_XPATH).to_s)
+  end
+
+  def tree_count
+    @decision_trees.count
+  end
+
+  def score(features)
+    formatted_features = Features.new(features).formatted
+    x = @decision_trees.map { |dt|
+      score = dt.decide(formatted_features)
+      score.to_s.to_f
+    }.reduce(:+) + @const
+    Math.exp(x) / (1 + Math.exp(x))
+  end
+
+end
+

data/lib/random_forest/node.rb

@@ -1,18 +1,23 @@
-require 'predicate'
+require 'simple_predicate'
+require 'simple_set_predicate'
 
 class Node
 
-  attr_reader :decision, :left, :right, :pred
+  attr_reader :decision, :pred, :children 
 
   def initialize(xml)
     children = xml.children
-    @pred = Predicate.new(children[0])
-
+    pred_xml = children[0]
+    @pred = SimplePredicate.new(pred_xml) if pred_xml.name == 'SimplePredicate'
+    @pred = SimpleSetPredicate.new(pred_xml) if pred_xml.name == 'SimpleSetPredicate'
+    @children = []
     @decision = xml.attribute('score').to_s
 
     return if children.count == 1
-    @left = Node.new(children[1]) if children[1]
-    @right = Node.new(children[2]) if children[2]
+
+    @children << Node.new(children[1]) if children[1]
+    @children << Node.new(children[2]) if children[2]
+    @children << Node.new(children[3]) if children[3]
   end
 
   def true?(features)

data/lib/random_forest/random_forest.rb

@@ -11,8 +11,9 @@ class RandomForest
   end
 
   def decisions_count(features)
+    formatted_features = Features.new(features).formatted
     decisions = @decision_trees.collect { |decision_tree|
-      decision_tree.decide(features)
+      decision_tree.decide(formatted_features)
     }
     decisions.inject(Hash.new(0)) { |h, e| h[e] += 1 ; h }
   end

data/lib/random_forest/simple_predicate.rb

@@ -20,9 +20,7 @@ class SimplePredicate
   end
 
   def true?(features)
-    format_boolean(features)
     return num_true?(features) if MATH_OPS.include?(@operator)
-
     return features[@field] == @value if @operator == EQUAL
     features[field].nil? || !features.has_key?(field) if @operator == IS_MISSING
   end
@@ -36,9 +34,4 @@ class SimplePredicate
     return curr_value <= value if @operator == LESS_OR_EQUAL
     curr_value >= value if @operator == GREATER_OR_EQUAL
   end
-
-  def format_boolean(features)
-    features[@field] = 'f' if features[@field] == false
-    features[@field] = 't' if features[@field] == true
-  end
 end

data/lib/random_forest/simple_set_predicate.rb

@@ -12,7 +12,6 @@ class SimpleSetPredicate
   end
 
   def true?(features)
-    format_boolean(features)
     @array.include? features[@field] if @operator == IS_IN
   end
 
@@ -23,9 +22,4 @@ class SimpleSetPredicate
         reject(&:empty?).
         map { |w| w.tr('"','')}
   end
-
-  def format_boolean(features)
-    features[@field] = 'f' if features[@field] == false
-    features[@field] = 't' if features[@field] == true
-  end
 end

data/lib/scoruby/version.rb

@@ -1,3 +1,3 @@
 module Scoruby
-  VERSION = '0.1.3'
+  VERSION = '0.2.0'
 end

data/scoruby.gemspec

@@ -9,7 +9,7 @@ Gem::Specification.new do |spec|
   spec.authors       = ["Asaf Schers"]
   spec.email         = ["schers@riskified.com"]
 
-  spec.summary       = %q{Creates a random forest object from a pmml file.}
+  spec.summary       = %q{Ruby Scoring API for PMML.}
   spec.homepage      = 'https://github.com/asafschers/scoruby'
   spec.license       = "MIT"
 
@@ -22,6 +22,7 @@ Gem::Specification.new do |spec|
   spec.add_development_dependency "rake", "~> 12.0"
   spec.add_development_dependency "rspec", "~> 3.5"
   spec.add_development_dependency "pry", "~> 0.10"
+  spec.add_development_dependency "coveralls"
   spec.add_development_dependency "ruby-prof"
   spec.add_dependency "nokogiri", "~> 1.7"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: scoruby
 version: !ruby/object:Gem::Version
-  version: 0.1.3
+  version: 0.2.0
 platform: ruby
 authors:
 - Asaf Schers
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2017-04-21 00:00:00.000000000 Z
+date: 2017-07-01 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -66,6 +66,20 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: '0.10'
+- !ruby/object:Gem::Dependency
+  name: coveralls
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
+  type: :development
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
 - !ruby/object:Gem::Dependency
   name: ruby-prof
   requirement: !ruby/object:Gem::Requirement
@@ -110,13 +124,10 @@ files:
 - Rakefile
 - bin/console
 - bin/setup
-- lib/gbm/gbm.rb
-- lib/gbm/gbm_decision_tree.rb
-- lib/gbm/gbm_node.rb
-- lib/gbm/gbm_predicate.rb
 - lib/random_forest/decision_tree.rb
+- lib/random_forest/features.rb
+- lib/random_forest/gbm.rb
 - lib/random_forest/node.rb
-- lib/random_forest/predicate.rb
 - lib/random_forest/random_forest.rb
 - lib/random_forest/simple_predicate.rb
 - lib/random_forest/simple_set_predicate.rb
@@ -148,6 +159,6 @@ rubyforge_project:
 rubygems_version: 2.2.2
 signing_key: 
 specification_version: 4
-summary: Creates a random forest object from a pmml file.
+summary: Ruby Scoring API for PMML.
 test_files: []
 has_rdoc: 

data/lib/gbm/gbm.rb

@@ -1,23 +0,0 @@
-require 'gbm_decision_tree'
-
-class Gbm
-  GBM_FOREST_XPATH = '//Segmentation[@multipleModelMethod="sum"]/Segment'
-  CONST_XPATH      = '//Constant[@dataType="double"]'
-
-  def initialize(xml)
-    @decision_trees = xml.xpath(GBM_FOREST_XPATH).collect{ |xml_tree|
-      GbmDecisionTree.new(xml_tree)
-    }
-    @const = Float(xml.xpath(CONST_XPATH).children[0].content)
-  end
-
-  def tree_count
-    @decision_trees.count
-  end
-
-  def score(features)
-    x = @decision_trees.map { |dt| dt.decide(features) }.reduce(:+) + @const
-    Math.exp(x) / (1 + Math.exp(x))
-  end
-end
-

data/lib/gbm/gbm_decision_tree.rb

@@ -1,37 +0,0 @@
-require 'gbm_node'
-
-class GbmDecisionTree
-  attr_reader :root
-
-  def initialize(tree_xml)
-    @id = tree_xml.attribute('id')
-    @root = GbmNode.new(tree_xml.xpath('TreeModel/Node'))
-  end
-
-  def decide(features)
-    curr = @root
-    while curr.score.nil?
-      prev = curr
-      curr = step(curr, features)
-      return if didnt_step?(curr, prev)
-    end
-
-    curr.score
-  end
-
-  private
-
-  def step(curr, features)
-    curr = curr.left if curr.left && curr.left.true?(features)
-    curr = curr.right if curr.right && curr.right.true?(features)
-    curr = curr.missing if curr.missing && curr.missing.true?(features)
-    curr
-  end
-
-  def didnt_step?(curr, prev)
-    return false if (prev.pred != curr.pred)
-    Scoruby.logger.error "Null tree: #{@id}, bad feature: #{curr.left.pred.field }"
-    true
-  end
-  
-end

data/lib/gbm/gbm_node.rb

@@ -1,23 +0,0 @@
-require 'gbm_predicate'
-
-class GbmNode
-
-  attr_reader :score, :missing, :left, :right, :pred
-
-  def initialize(xml)
-    children = xml.children
-    @pred = GbmPredicate.new(children[0])
-
-    @score = xml.attribute('score').to_s.to_f unless xml.attribute('score').to_s.empty?
-
-    return if children.count == 1
-    @missing = GbmNode.new(children[1]) if children[1]
-    @left = GbmNode.new(children[2]) if children[2]
-    @right = GbmNode.new(children[3]) if children[3]
-  end
-
-  def true?(features)
-    @pred.nil? || @pred.true?(features)
-  end
-
-end

data/lib/gbm/gbm_predicate.rb

@@ -1,18 +0,0 @@
-require 'simple_predicate'
-require 'simple_set_predicate'
-
-class GbmPredicate
-
-  def initialize(pred_xml)
-    @pred = SimplePredicate.new(pred_xml) if pred_xml.name == 'SimplePredicate'
-    @pred = SimpleSetPredicate.new(pred_xml) if pred_xml.name == 'SimpleSetPredicate'
-  end
-
-  def field
-    @pred.field
-  end
-
-  def true?(features)
-    @pred.true?(features)
-  end
-end

data/lib/random_forest/predicate.rb

@@ -1,32 +0,0 @@
-require 'simple_predicate'
-require 'simple_set_predicate'
-
-class Predicate
-
-  def initialize(pred_xml)
-    @pred = SimplePredicate.new(pred_xml) if pred_xml.name == 'SimplePredicate'
-    @pred = SimpleSetPredicate.new(pred_xml) if pred_xml.name == 'SimpleSetPredicate'
-  end
-
-  def field
-    @pred.field
-  end
-
-  def true?(features)
-    return if missing_feature?(features)
-    return if nil_feature?(features)
-    @pred.true?(features)
-  end
-
-  def missing_feature?(features)
-    return false if features.has_key? field
-    Scoruby.logger.error "Missing feature #{field}"
-    true
-  end
-
-  def nil_feature?(features)
-    return false unless features[field].nil?
-    Scoruby.logger.error "Feature #{field} value is nil"
-    true
-  end
-end