scoruby 0.2.4 → 0.2.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 5ab6779638032408bc0a38812ded29891831aebd
-  data.tar.gz: d26d4338fdbb2f00a7d6fed974a6e2d577c1d847
+  metadata.gz: 6d01f41ef6e3e3485acb65587892402866004bd2
+  data.tar.gz: 7d898f2a4c86915e19952a3641286dd252d7c3fd
 SHA512:
-  metadata.gz: 080081fe11bdba935bdd99b178ce74a04d1d05dadc080283f9e474a4d84e8415d2a39cf6a9916c4ee46c78fb46965aa934a0850491de04f6fd1d5806aaa32463
-  data.tar.gz: 38a7ac8547797f2367fcd66036ea87971c531106c0dd0b4c56abf684f48a72db14a9139fa5ff29b0e809aadb5ffb0908a265a603aa6d3ddf5725dc4858cdab0a
+  metadata.gz: de6e7acbbcf5acd97f2b980253be94c7eb2fb7c31fb76c9c9d1fe6ca09447751974734370826d1477db639685a6ab37108ced5d85e4fc9b397285e1b61468a4e
+  data.tar.gz: 92e5728b151d7c3daa24e5fccebe6b7e8567c92af4fb03cdbc5b5ef20f5a75b53c88b5ef79e6d4db20df157995f45b6222677a0e1634702e3769240b61f025d4

data/.gitignore

@@ -3,12 +3,12 @@ coverage
 
 *.log
 
+sample.pmml
+
 *.gem
 
 spec/fixtures/decision_tree_v2.pmml
 
-sample.pmml
-
 test_gbm.pmml
 
 test_gbm.rb

data/.travis.yml

@@ -1,4 +1,4 @@
 language: ruby
 rvm:
-  - 2.1.2
+  - 2.4.1
 before_install: gem install bundler -v 1.10.5

data/README.md

@@ -7,7 +7,7 @@
 
 Ruby scoring API for Predictive Model Markup Language (PMML).
 
-Currently supports random forest and gradient boosted models.
+Currently supports Decision Tree, Random Forest and Gradient Boosted Models.
 
 Will be happy to implement new models by demand, or assist with any other issue.
 
@@ -31,37 +31,8 @@ Or install it yourself as:
 
 ## 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')
-
-```
 
+[Generate PMML - R](https://github.com/asafschers/scoruby/wiki/Random-Forest) 
 #### Classify by PMML - Ruby
 
 ```ruby
@@ -89,34 +60,7 @@ random_forest.decisions_count(features)
 
 ### 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')
-
-```
+[Generate PMML - R](https://github.com/asafschers/scoruby/wiki/Gradient-Boosted-Model) 
 
 #### Classify by PMML - Ruby
 
@@ -142,8 +86,6 @@ gbm.score(features)
 
 ### Decision Tree
 
-#### Classify by PMML - Ruby
-
 ```ruby
 decision_tree = Scoruby.get_model 'decision_tree.pmml'
 

data/lib/scoruby/models/naive_bayes.rb

@@ -0,0 +1,92 @@
+module Scoruby
+  module Models
+    class NaiveBayes
+      attr_reader :data
+
+      def initialize(xml)
+        @threshold = xml.xpath('//NaiveBayesModel').attr('threshold').value.to_f
+        @data = {}
+        xml.xpath('//BayesInput').each do |feature|
+          @data[feature.attr('fieldName').to_sym] = fetch_feature(feature)
+        end
+
+        @labels = {}
+        xml.xpath('//BayesOutput//TargetValueCount').each do |l| l.attr('value')
+          @labels[l.attr('value')] = { 'count': l.attr('count').to_f }
+        end
+      end
+
+      def lvalues(features)
+        @labels.each do |label, _|
+          features.each do |feature_name, feature_value|
+
+            if @data[feature_name][feature_value]
+              value_count = @data[feature_name][feature_value][label].to_f
+              overall_count = @data[feature_name].sum { |_, value| value[label].to_f }
+
+              @labels[label][feature_name] = value_count / overall_count
+            elsif @data[feature_name][label]
+              @labels[label][feature_name] = calc_numerical(@data[feature_name][label], feature_value)
+            end
+          end
+        end
+
+        lvalues = {}
+        @labels.each do |label, label_data|
+          label_data.each do |key, value|
+            label_data[key] = @threshold if value.round(5).zero?
+          end
+          lvalues[label] = label_data.values.reduce(:*)
+        end
+        lvalues
+      end
+
+      def score(features, label)
+        lvalues = lvalues(features)
+        lvalues[label] / lvalues.values.reduce(:+)
+      end
+
+      private
+
+      def calc_numerical(label_data, feature_value)
+        variance = label_data[:variance].to_f
+        mean = label_data[:mean].to_f
+        feature_value = feature_value.to_f
+
+        Math.exp(-(feature_value - mean)**2 / (2 * variance)) / Math.sqrt(2 * Math::PI * variance)
+      end
+
+      def fetch_feature(feature)
+        return fetch_numerical_feature(feature) if feature.child.name == 'TargetValueStats'
+        fetch_category_feature(feature)
+      end
+
+      def fetch_numerical_feature(feature)
+        features_data = {}
+        feature.child.children.each do |child|
+          features_data[child.attr('value').strip] = {
+            mean: child.child.attr('mean'),
+            variance: child.child.attr('variance')
+          }
+        end
+        features_data
+      end
+
+      def fetch_category_feature(feature)
+        feature_data = {}
+        feature.children.each do |category|
+          feature_data[category.attr('value')] = fetch_category(category)
+        end
+        feature_data
+      end
+
+      def fetch_category(category)
+        category_data = {}
+        category.child.children.each do |label|
+          category_data[label.attr('value')] = label.attr('count')
+        end
+        category_data
+      end
+    end
+  end
+end

data/lib/scoruby/version.rb

@@ -1,3 +1,3 @@
 module Scoruby
-  VERSION = '0.2.4'
+  VERSION = '0.2.5'
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: scoruby
 version: !ruby/object:Gem::Version
-  version: 0.2.4
+  version: 0.2.5
 platform: ruby
 authors:
 - Asaf Schers
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2017-07-22 00:00:00.000000000 Z
+date: 2017-09-14 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -131,6 +131,7 @@ files:
 - lib/scoruby/features.rb
 - lib/scoruby/models/decision_tree.rb
 - lib/scoruby/models/gbm.rb
+- lib/scoruby/models/naive_bayes.rb
 - lib/scoruby/models/random_forest.rb
 - lib/scoruby/models_factory.rb
 - lib/scoruby/node.rb