naivebayes 0.0.3 → 0.1.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 6bbec8539369ff2f707fd59440b444cbd2a71e9f
-  data.tar.gz: 8b0fa15a8dec7d4e314c60cff086074f48f37115
+  metadata.gz: ba428b87232e40d36c9c97f615a0e865bef57dfc
+  data.tar.gz: 8e195ed2e429a661743dba99236efe6f1529c706
 SHA512:
-  metadata.gz: a3210aff3d52771134877dd65b4a9cf524dfd6a754f523eb708b85da4e651e00435b97be6fcca66e63fb2aa37313706012eab99405e940f2d20cd316e8ec43cc
-  data.tar.gz: d25b10280650406e9e33566641d9c5c9846a59a2fd81f8da9d38d147774174d348e2a43624946154b962c6b6aa98bade8aaebd2c35fcafb3cc54f4f6c075139f
+  metadata.gz: 9ca9adf60295e6765bbec1968b486592c5cdc855bc868e0b8a4de982a28c91537985b3521bf28992377f8e03714218db48af7857093ed2dde3e5eeda490cae27
+  data.tar.gz: 64e04a038114c4bef08073f55320b7291ee5edb1327e6f2caed868de9edc4559154e87159ebaf125c5c9078d8fd19a6a3736fff624cfa5e46039a0f5486789b9

data/README.md

@@ -40,6 +40,20 @@ result = classifier.classify({"aaa" => 1, "bbb" => 1})
 p result # => {"positive" => 0.9411764705882353,"negative" => 0.05882352941176469}
 ```
 
+Complement Naive Bayes.
+
++ http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.13.8572
+
+``` html
+require 'naivebayes'
+classifier = NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 1)
+classifier.train("positive", {"aaa" => 3, "bbb" => 1, "ccc" => 2})
+classifier.train("negative", {"aaa" => 1, "bbb" => 4, "ccc" => 2})
+classifier.train("neutral",  {"aaa" => 2, "bbb" => 3, "ccc" => 5})
+result = classifier.classify({"aaa" => 4, "bbb" => 3, "ccc" => 3})
+p result #=> {"neutral"=>9.985931139006835, "negative"=>10.112101263742268, "positive"=>10.836883752313222}
+```
+
 
 ChangeLog
 ---------

data/VERSION

@@ -1 +1 @@
-0.0.3
+0.1.0

data/demo.rb

@@ -23,3 +23,13 @@ classifier.train("negative", {"ccc" => 2, "ddd" => 3})
 result = classifier.classify({"aaa" => 1, "bbb" => 1})
 
 p result # => {"positive" => 0.9411764705882353,"negative" => 0.05882352941176469}
+
+classifier = NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 1)
+
+classifier.train("positive", {"aaa" => 3, "bbb" => 1, "ccc" => 2})
+classifier.train("negative", {"aaa" => 1, "bbb" => 4, "ccc" => 2})
+classifier.train("neutral",  {"aaa" => 2, "bbb" => 3, "ccc" => 5})
+result = classifier.classify({"aaa" => 4, "bbb" => 3, "ccc" => 3})
+
+p result #=> {"neutral"=>9.985931139006835, "negative"=>10.112101263742268, "positive"=>10.836883752313222}
+

data/doc/ChangeLog

@@ -1,3 +1,8 @@
+=== 0.1.0 / 2014-01-23
+
+* Implement Complement Naive Bayes.
+
+
 === 0.0.3 / 2013-12-13
 
 * Accessable instance variable.

data/lib/naivebayes/classifier.rb

@@ -3,7 +3,7 @@
 
 module NaiveBayes
   class Classifier
-    attr_accessor :frequency_table, :word_table, :instance_count_of, :total_count, :model
+    attr_accessor :frequency_table, :word_table, :instance_count_of, :total_count, :model, :smoothing_parameter
 
     def initialize(params = {})
       @frequency_table = Hash.new
@@ -11,17 +11,18 @@ module NaiveBayes
       @instance_count_of = Hash.new(0)
       @total_count = 0
       @model = params[:model]
+      @smoothing_parameter = params[:smoothing_parameter] || 1
     end
 
-    def train(label, attributes)
+    def train(label, feature)
       unless @frequency_table.has_key?(label)
         @frequency_table[label] = Hash.new(0)
       end
-      attributes.each {|word, frequency|
-        if @model == "multinomial"
-          @frequency_table[label][word] += frequency
-        else
+      feature.each {|word, frequency|
+        if @model == "berounoulli"
           @frequency_table[label][word] += 1
+        else
+          @frequency_table[label][word] += frequency
         end
         @word_table[word] = 1
       }
@@ -29,7 +30,13 @@ module NaiveBayes
       @total_count += 1
     end
 
-    def classify(attributes)
+    def classify(feature)
+      @model == "complement" ? cnb(feature) : mnb(feature)
+    end
+
+    private
+
+    def mnb(feature)
       class_prior_of = Hash.new(1)
       likelihood_of = Hash.new(1)
       class_posterior_of = Hash.new(1)
@@ -42,7 +49,7 @@ module NaiveBayes
         @word_table.each_key {|word|
           laplace_word_likelihood = (@frequency_table[label][word] + 1).to_f /
             (@instance_count_of[label] + @word_table.size()).to_f
-          if attributes.has_key?(word)
+          if feature.has_key?(word)
             likelihood_of[label] *= laplace_word_likelihood
           else
             likelihood_of[label] *= (1 - laplace_word_likelihood)
@@ -56,5 +63,37 @@ module NaiveBayes
       }
       return class_posterior_of
     end
+
+    def cnb(feature)
+      all_class = @frequency_table.keys
+      all_train_data = @instance_count_of.values.inject(0) {|s, v| s + v}
+      class_posterior_of = all_class.map {|c|
+        n_c = total_number_of_word_in_other_class(c)
+        alpha = @smoothing_parameter*feature.length
+        term2nd = feature.to_a.map {|e|
+          k = e[0]
+          v = e[1]
+          v*Math.log((number_of_word_in_other_class(c, k) + @smoothing_parameter).to_f/(n_c + alpha))
+        }.inject(0) {|s, v| s + v}
+        theta_c = @instance_count_of[c].to_f/all_train_data
+        [c, Math.log(theta_c) - term2nd]
+      }.sort {|x, y| x[1] <=> y[1]}.flatten
+      Hash[*class_posterior_of]
+    end
+
+    def total_number_of_word_in_other_class(c)
+      all_words = @frequency_table.values.map {|h| h.keys}.flatten.sort.uniq
+      other_classes = @frequency_table.keys - [c]
+      other_classes.map {|c|
+        all_words.map {|w|
+          @frequency_table[c][w]
+        }
+      }.flatten.inject(0) {|s, v| s + v}
+    end
+
+    def number_of_word_in_other_class(c, i)
+      other_classes = @frequency_table.keys - [c]
+      other_classes.map {|c| @frequency_table[c][i]}.inject(0) {|s, v| s + v}
+    end
   end
 end

data/lib/naivebayes/version.rb

@@ -1,10 +1,10 @@
 # Name::      NaiveBayes::Version
 # Author::    774 <http://id774.net>
 # Created::   Nov 24, 2013
-# Updated::   Dec 13, 2013
-# Copyright:: 774 Copyright (c) 2013
+# Updated::   Jan 22, 2014
+# Copyright:: 774 Copyright (c) 2013-2014
 # License::   Licensed under the GNU GENERAL PUBLIC LICENSE, Version 3.0.
 
 module NaiveBayes
-  VERSION = "0.0.3"
+  VERSION = "0.1.0"
 end

data/naivebayes.gemspec

@@ -2,15 +2,16 @@
 # DO NOT EDIT THIS FILE DIRECTLY
 # Instead, edit Jeweler::Tasks in Rakefile, and run 'rake gemspec'
 # -*- encoding: utf-8 -*-
-# stub: naivebayes 0.0.3 ruby lib
+# stub: naivebayes 0.1.0 ruby lib
 
 Gem::Specification.new do |s|
   s.name = "naivebayes"
-  s.version = "0.0.3"
+  s.version = "0.1.0"
 
   s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
+  s.require_paths = ["lib"]
   s.authors = ["id774"]
-  s.date = "2013-12-13"
+  s.date = "2014-01-23"
   s.description = "Naive Bayes classifier"
   s.email = "idnanashi@gmail.com"
   s.extra_rdoc_files = [
@@ -40,8 +41,7 @@ Gem::Specification.new do |s|
   ]
   s.homepage = "http://github.com/id774/naivebayes"
   s.licenses = ["GPL"]
-  s.require_paths = ["lib"]
-  s.rubygems_version = "2.1.11"
+  s.rubygems_version = "2.2.0"
   s.summary = "naivebayes"
 
   if s.respond_to? :specification_version then

data/spec/lib/naivebayes/classifier_spec.rb

@@ -180,6 +180,28 @@ describe NaiveBayes::Classifier do
         expect(subject).to eq expected
       end
     end
+
+    context '@model with complement' do
+      subject { classifier.model }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return model name' do
+        expected = "complement"
+        expect(subject).to eq expected
+      end
+    end
+
+    context '@smoothing_parameter with complement' do
+      subject { classifier.smoothing_parameter }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 3) }
+
+      it 'should return smoothing parameter' do
+        expected = 3
+        expect(subject).to eq expected
+      end
+    end
   end
 end
 
@@ -267,6 +289,106 @@ describe NaiveBayes::Classifier do
   end
 end
 
+describe NaiveBayes::Classifier do
+  describe 'Complement Naive Bayes' do
+    context 'with train data of two expecting positive' do
+
+      subject { classifier.classify({"aaa" => 1, "bbb" => 1}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return positive' do
+        classifier.train("positive", {"aaa" => 0, "bbb" => 1})
+        classifier.train("negative", {"ccc" => 2, "ddd" => 3})
+
+        expected = {
+          "negative"=>0.8109302162163289,
+          "positive"=>3.1986731175506815
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of two expecting negative' do
+
+      subject { classifier.classify({"ccc" => 3, "ddd" => 3}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return negative' do
+        classifier.train("positive", {"aaa" => 0, "bbb" => 1})
+        classifier.train("negative", {"ccc" => 2, "ddd" => 3})
+
+        expected = {
+          "positive"=>3.527593764407934,
+          "negative"=>5.898526551448713
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of two expecting negative and smoothing parameter 1' do
+
+      subject { classifier.classify({"ccc" => 3, "ddd" => 3}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 1) }
+
+      it 'should return negative' do
+        classifier.train("positive", {"aaa" => 0, "bbb" => 1})
+        classifier.train("negative", {"ccc" => 2, "ddd" => 3})
+
+        expected = {
+          "positive"=>3.527593764407934,
+          "negative"=>5.898526551448713
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of two expecting positive and smoothing parameter 3' do
+
+      subject { classifier.classify({"aaa" => 1, "bbb" => 1}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 3) }
+
+      it 'should return positive' do
+        classifier.train("positive", {"aaa" => 0, "bbb" => 1})
+        classifier.train("negative", {"ccc" => 2, "ddd" => 3})
+
+        expected = {
+          "negative"=>0.7137664677626813,
+          "positive"=>1.9054187877005764
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of two expecting positive and smoothing parameter 5' do
+
+      subject { classifier.classify({"aaa" => 1, "bbb" => 1}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement", :smoothing_parameter => 5) }
+
+      it 'should return positive' do
+        classifier.train("positive", {"aaa" => 0, "bbb" => 1})
+        classifier.train("negative", {"ccc" => 2, "ddd" => 3})
+
+        expected = {
+          "negative"=>0.7014459833746406,
+          "positive"=>1.5040773967762742
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+  end
+end
+
 describe NaiveBayes::Classifier do
   describe 'The berounoulli model' do
     context 'with train data of three expecting positive' do
@@ -400,3 +522,70 @@ describe NaiveBayes::Classifier do
     end
   end
 end
+
+describe NaiveBayes::Classifier do
+  describe 'Complement Naive Bayes' do
+    context 'with train data of three expecting positive' do
+
+      subject { classifier.classify({"aaa" => 4, "bbb" => 3, "ccc" => 3}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return positive' do
+        classifier.train("positive", {"aaa" => 3, "bbb" => 1, "ccc" => 2})
+        classifier.train("negative", {"aaa" => 1, "bbb" => 4, "ccc" => 2})
+        classifier.train("neutral",  {"aaa" => 2, "bbb" => 3, "ccc" => 5})
+
+        expected = {
+          "neutral"  => 9.985931139006835,
+          "negative" => 10.112101263742268,
+          "positive" => 10.836883752313222
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of three expecting negative' do
+
+      subject { classifier.classify({"aaa" => 3, "bbb" => 4, "ccc" => 3}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return negative' do
+        classifier.train("positive", {"aaa" => 3, "bbb" => 1, "ccc" => 2})
+        classifier.train("negative", {"aaa" => 1, "bbb" => 4, "ccc" => 2})
+        classifier.train("neutral",  {"aaa" => 2, "bbb" => 3, "ccc" => 5})
+
+        expected = {
+          "neutral"  => 9.80360958221288,
+          "positive" => 10.143736571753276,
+          "negative" => 10.294422820536223
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+
+    context 'with train data of three expecting neutral' do
+
+      subject { classifier.classify({"aaa" => 3, "bbb" => 3, "ccc" => 5}) }
+
+      let(:classifier) { NaiveBayes::Classifier.new(:model => "complement") }
+
+      it 'should return neutral' do
+        classifier.train("positive", {"aaa" => 3, "bbb" => 1, "ccc" => 2})
+        classifier.train("negative", {"aaa" => 1, "bbb" => 4, "ccc" => 2})
+        classifier.train("neutral",  {"aaa" => 2, "bbb" => 3, "ccc" => 5})
+
+        expected = {
+          "negative" => 10.68941662877709,
+          "positive" => 11.06002730362743,
+          "neutral"  => 11.149081948812517
+        }
+
+        expect(subject).to eq expected
+      end
+    end
+  end
+end

data/spec/lib/naivebayes_spec.rb

@@ -6,6 +6,6 @@ describe NaiveBayes do
   context "VERSION" do
     subject { NaiveBayes::VERSION }
 
-    it { expect(subject).to eq "0.0.3" }
+    it { expect(subject).to eq "0.1.0" }
   end
 end

metadata

@@ -1,55 +1,55 @@
 --- !ruby/object:Gem::Specification
 name: naivebayes
 version: !ruby/object:Gem::Version
-  version: 0.0.3
+  version: 0.1.0
 platform: ruby
 authors:
 - id774
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2013-12-13 00:00:00.000000000 Z
+date: 2014-01-23 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: cucumber
   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: bundler
   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: jeweler
   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'
 description: Naive Bayes classifier
@@ -89,17 +89,17 @@ require_paths:
 - lib
 required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
-  - - '>='
+  - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
-  - - '>='
+  - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
 rubyforge_project: 
-rubygems_version: 2.1.11
+rubygems_version: 2.2.0
 signing_key: 
 specification_version: 4
 summary: naivebayes