bayes_motel 0.1.0

data/.document

@@ -0,0 +1,5 @@
+README.rdoc
+lib/**/*.rb
+bin/*
+features/**/*.feature
+LICENSE

data/.gitignore

@@ -0,0 +1,21 @@
+## MAC OS
+.DS_Store
+
+## TEXTMATE
+*.tmproj
+tmtags
+
+## EMACS
+*~
+\#*
+.\#*
+
+## VIM
+*.swp
+
+## PROJECT::GENERAL
+coverage
+rdoc
+pkg
+
+## PROJECT::SPECIFIC

data/LICENSE

@@ -0,0 +1,20 @@
+Copyright (c) 2010 Mike Perham
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

data/README.md

@@ -0,0 +1,55 @@
+bayes_motel
+--------------
+
+BayesMotel is a multi-variate Bayesian classification engine.  There are two steps to Bayesian classification:
+
+1. Training
+You provide a set of variables along with the proper classification for that set.
+2. Runtime
+You provide a set of variables and ask for the proper classification according to the training in Step 1.
+
+Commonly this is used for spam detection.  You will provide a corpus of emails or other data along with a "Spam/NotSpam" classification.  The library will determine which variables affect the classification and use that to judge future data.
+
+
+Usage
+=============
+
+Step one is to create a corpus that you can train with a set of previously classified documents:
+
+    corpse = BayesMotel::Corpus.new('tweets')
+    spam_tweets.each do |tweet|
+      corpse.train(tweet, :spam)
+    end
+    good_tweets.each do |tweet|
+      corpse.train(tweet, :ham)
+    end
+    corpse.cleanup
+
+In this example, we have a set of spammy tweets and a set of known good tweets.  We pass in each tweet
+to our train() method.  Once we have completed training, we call cleanup which will run through the
+internal data structures and clean up any variables that are too 'unique' to make a difference in classification (for instance, an :id variable will be unique for each tweet and so will be removed in the cleanup since it does not repeat enough times).
+
+Step two is to use the calculated corpus for the category scores or a classification for a given document:
+
+    corpse.scores(new_tweet)
+    => { :spam => 12.4, :ham => 15.25 }
+    corpse.classify(new_tweet)
+    => [:ham, 15.25]
+
+
+Trivia
+==============
+
+Bates Motel is the motel in Alfred Hitchcock's masterpiece _Psycho_.  Corpus is Latin for "body" but also means 'a canonical set of documents'.  I'm not crazy, I just like puns.
+
+
+Author
+==============
+
+Mike Perham, mperham AT gmail.com, @mperham, http://mikeperham.com
+
+
+Copyright
+==============
+
+Copyright (c) 2010 Mike Perham. See LICENSE for details.

data/Rakefile

@@ -0,0 +1,53 @@
+require 'rubygems'
+require 'rake'
+require 'lib/bayes_motel/version'
+
+begin
+  require 'jeweler'
+  Jeweler::Tasks.new do |gem|
+    gem.name = "bayes_motel"
+    gem.summary = %Q{Bayesian classification engine}
+    gem.description = %Q{http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/}
+    gem.email = "mperham@gmail.com"
+    gem.homepage = "http://github.com/mperham/bayes_motel"
+    gem.authors = ["Mike Perham"]
+    gem.version = BayesMotel::VERSION
+    gem.add_development_dependency "shoulda", ">= 0"
+    # gem is a Gem::Specification... see http://www.rubygems.org/read/chapter/20 for additional settings
+  end
+  Jeweler::GemcutterTasks.new
+rescue LoadError
+  puts "Jeweler (or a dependency) not available. Install it with: gem install jeweler"
+end
+
+require 'rake/testtask'
+Rake::TestTask.new(:test) do |test|
+  test.libs << 'lib' << 'test'
+  test.pattern = 'test/**/test_*.rb'
+  test.verbose = true
+end
+
+begin
+  require 'rcov/rcovtask'
+  Rcov::RcovTask.new do |test|
+    test.libs << 'test'
+    test.pattern = 'test/**/test_*.rb'
+    test.verbose = true
+  end
+rescue LoadError
+  task :rcov do
+    abort "RCov is not available. In order to run rcov, you must: sudo gem install spicycode-rcov"
+  end
+end
+
+task :test => :check_dependencies
+
+task :default => :test
+
+require 'rake/rdoctask'
+Rake::RDocTask.new do |rdoc|
+  rdoc.rdoc_dir = 'rdoc'
+  rdoc.title = "bayes_motel #{BayesMotel::VERSION}"
+  rdoc.rdoc_files.include('README*')
+  rdoc.rdoc_files.include('lib/**/*.rb')
+end

data/lib/bayes_motel.rb

@@ -0,0 +1,3 @@
+require 'bayes_motel/version'
+require 'bayes_motel/persistence'
+require 'bayes_motel/corpus'

data/lib/bayes_motel/corpus.rb

@@ -0,0 +1,97 @@
+module BayesMotel
+  class Corpus
+    attr_reader :name
+    attr_reader :total_count
+    attr_reader :data
+
+    def initialize(name)
+      @name = name
+      @total_count = 0
+      @data = {}
+    end
+    
+    def train(doc, category)
+      @total_count += 1
+      _training(doc, category)
+    end
+    
+    def score(doc)
+      _score(doc)
+    end
+      
+    # The default classification algorithm just picks
+    # the category with the highest score.
+    def classify(doc)
+      results = score(doc)
+      max = [:none, 0]
+      results.each_pair do |(k, v)|
+        max = [k, v] if v > max[1]
+      end
+      max
+    end
+    
+    def cleanup
+      @data.each_pair do |k, v|
+        clean(@data, k, v)
+      end
+    end
+    
+    private
+
+    def _score(variables, name='', odds={})
+      variables.each_pair do |k, v|
+        case v
+        when Hash
+          _score(v, "#{name}_#{k}", odds)
+        else
+          @data.each_pair do |category, raw_counts|
+            cat = odds[category] ||= {}
+            keys = raw_counts["#{name}_#{k}"] || {}
+            cat["#{name}_#{k}_#{v}"] = Float(keys[v] || 0) / @total_count
+          end
+        end
+      end
+      odds.inject({}) do |memo, (key, value)|
+        memo[key] = value.inject(0) do |acc_memo, (acc_key, acc_value)|
+          acc_memo += acc_value
+        end
+        memo
+      end
+    end
+    
+    def _training(variables, category, name='')
+      variables.each_pair do |k, v|
+        case v
+        when Hash
+          _training(v, category, "#{name}_#{k}")
+        else
+          cat = (@data[category] ||= {})
+          values = (cat["#{name}_#{k}"] ||= {})
+          values[v] ||= 0
+          values[v] += 1
+        end
+      end
+    end
+
+    def clean(hash, k, v)
+      case v
+      when Hash
+        v.each_pair do |key, value|
+          clean(v, key, value)
+        end
+        if v.empty?
+          hash.delete(k)
+        elsif v.size == 1 and v['other']
+          hash.delete(k)
+        end
+      else
+        if v < (@total_count * 0.03).floor
+          hash['other'] ||= 0
+          hash['other'] += v
+          hash.delete(k)
+        end
+      end
+    end
+
+  end
+end

data/lib/bayes_motel/persistence.rb

@@ -0,0 +1,14 @@
+module BayesMotel
+  module Persistence
+    # TODO Make this a little more Ruby idiomatic and pluggable
+    # for filesystems, databases, etc.
+    def self.write(corpus)
+      File.open("#{corpus.name}", 'w') do |file|
+        Marshal.dump(corpus, file)
+      end
+    end
+    def self.read(name)
+      Marshal.load(File.read("#{name}"))
+    end
+  end
+end

data/lib/bayes_motel/version.rb

@@ -0,0 +1,3 @@
+module BayesMotel
+  VERSION = "0.1.0"
+end

data/test/helper.rb

@@ -0,0 +1,11 @@
+require 'rubygems'
+require 'test/unit'
+require 'shoulda'
+require 'zlib'
+
+$LOAD_PATH.unshift(File.join(File.dirname(__FILE__), '..', 'lib'))
+$LOAD_PATH.unshift(File.dirname(__FILE__))
+require 'bayes_motel'
+
+class Test::Unit::TestCase
+end

data/test/test_bayes_motel.rb

@@ -0,0 +1,38 @@
+require 'helper'
+
+class TestBayesMotel < Test::Unit::TestCase
+  
+  should "allow basic training" do
+    c = BayesMotel::Corpus.new('email')
+    tweets.each do |tweet|
+      c.train(tweet, :ham)
+    end
+    c.cleanup
+    assert_equal tweets.size, c.total_count
+  end
+  
+  should "allow big training" do
+    c = BayesMotel::Corpus.new('email')
+    tweets(2000).each do |tweet|
+      c.train(tweet, :ham)
+    end
+    c.cleanup
+    assert_equal tweets.size, c.total_count
+  end
+  
+  private
+  
+  def tweets(n=100)
+    @tweets ||= begin
+      t = []
+      Zlib::GzipReader.open("#{File.dirname(__FILE__)}/#{n}tweets.txt.gz") do |gz|
+        gz.read.each_line do |line|
+          hash = eval(line)
+          hash.delete(:retweeted_status) if hash[:retweeted_status]
+          t << hash
+        end
+      end
+      t
+    end
+  end
+end

data/test/test_persistence.rb

@@ -0,0 +1,22 @@
+require 'helper'
+require 'fileutils'
+
+class TestPersistence < Test::Unit::TestCase
+
+  should "persist" do
+    c1 = BayesMotel::Corpus.new('test1')
+    c1.train({ :something => 'foo', :kids => { :bar => 'whiz', :id => 123 } }, :ham)
+    c1.train({ :something => 'foo', :kids => { :bar => 'gee', :id => 145 } }, :spam)
+    
+    BayesMotel::Persistence.write(c1)
+    c2 = BayesMotel::Persistence.read('test1')
+    FileUtils.rm_f 'test1'
+    assert c1 != c2
+    c1.instance_variables.each do |var|
+      v1 = c1.instance_variable_get(var)
+      v2 = c2.instance_variable_get(var)
+      assert_equal v1, v2
+    end
+  end
+
+end

data/test/test_training.rb

@@ -0,0 +1,54 @@
+require 'helper'
+
+class TestTraining < Test::Unit::TestCase
+  
+  should "handle basic training" do
+    c = BayesMotel::Corpus.new('test')
+    c.train({ :something => 'foo' }, :ham)
+    c.train({ :something => 'foo' }, :spam)
+    
+    results = c.score({ :something => 'foo' })
+    assert results
+    assert_equal 2, results.size
+    assert_equal results[:spam], results[:ham]
+  end
+
+  should "not care about extra variables" do
+    c = BayesMotel::Corpus.new('test')
+    c.train({ :something => 'foo' }, :ham)
+    c.train({ :something => 'foo', :fubwhiz => 'oh noes' }, :spam)
+    
+    results = c.score({ :something => 'foo' })
+    assert results
+    assert_equal 2, results.size
+    assert_equal results[:spam], results[:ham]
+  end
+
+  should "give more weight with more appearances" do
+    c = BayesMotel::Corpus.new('test')
+    c.train({ :something => 'foo' }, :ham)
+    c.train({ :something => 'foo' }, :spam)
+    c.train({ :something => 'foo', :fubwhiz => 'oh noes' }, :spam)
+    
+    doc = { :something => 'foo' }
+    results = c.score(doc)
+    assert results
+    assert_equal 2, results.size
+    assert_equal results[:spam], 2*results[:ham]
+    assert_equal [:spam, 2.0/3], c.classify(doc)
+  end
+  
+  should "calculate score for nested documents" do
+    c = BayesMotel::Corpus.new('test')
+    c.train({ :something => 'foo', :kids => { :bar => 'whiz', :id => 123 } }, :ham)
+    c.train({ :something => 'foo', :kids => { :bar => 'gee', :id => 145 } }, :spam)
+
+    doc = { :something => 'foo', :kids => { :bar => 'gee', :id => 167, :ack => 'blag' }}
+    results = c.score(doc)
+    assert results
+    assert_equal 2, results.size
+    assert_equal results[:spam], 2*results[:ham]
+    
+    assert_equal [:spam, 1.0], c.classify(doc)
+  end
+end

metadata

@@ -0,0 +1,91 @@
+--- !ruby/object:Gem::Specification 
+name: bayes_motel
+version: !ruby/object:Gem::Version 
+  prerelease: false
+  segments: 
+  - 0
+  - 1
+  - 0
+  version: 0.1.0
+platform: ruby
+authors: 
+- Mike Perham
+autorequire: 
+bindir: bin
+cert_chain: []
+
+date: 2010-04-28 00:00:00 -05:00
+default_executable: 
+dependencies: 
+- !ruby/object:Gem::Dependency 
+  name: shoulda
+  prerelease: false
+  requirement: &id001 !ruby/object:Gem::Requirement 
+    requirements: 
+    - - ">="
+      - !ruby/object:Gem::Version 
+        segments: 
+        - 0
+        version: "0"
+  type: :development
+  version_requirements: *id001
+description: http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/
+email: mperham@gmail.com
+executables: []
+
+extensions: []
+
+extra_rdoc_files: 
+- LICENSE
+- README.md
+files: 
+- .document
+- .gitignore
+- LICENSE
+- README.md
+- Rakefile
+- lib/bayes_motel.rb
+- lib/bayes_motel/corpus.rb
+- lib/bayes_motel/persistence.rb
+- lib/bayes_motel/version.rb
+- test/100tweets.txt.gz
+- test/2000tweets.txt.gz
+- test/helper.rb
+- test/test_bayes_motel.rb
+- test/test_persistence.rb
+- test/test_training.rb
+has_rdoc: true
+homepage: http://github.com/mperham/bayes_motel
+licenses: []
+
+post_install_message: 
+rdoc_options: 
+- --charset=UTF-8
+require_paths: 
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement 
+  requirements: 
+  - - ">="
+    - !ruby/object:Gem::Version 
+      segments: 
+      - 0
+      version: "0"
+required_rubygems_version: !ruby/object:Gem::Requirement 
+  requirements: 
+  - - ">="
+    - !ruby/object:Gem::Version 
+      segments: 
+      - 0
+      version: "0"
+requirements: []
+
+rubyforge_project: 
+rubygems_version: 1.3.6
+signing_key: 
+specification_version: 3
+summary: Bayesian classification engine
+test_files: 
+- test/helper.rb
+- test/test_bayes_motel.rb
+- test/test_persistence.rb
+- test/test_training.rb