coletivo 0.0.1 → 0.0.2

data/README.rdoc

@@ -1,6 +1,7 @@
 = coletivo
 
-An awesome, flexible, powerful, useful, tricky and liar Rails 3 recommendations engine.
+A simple Rails 3 recommendations engine.
+Coletivo uses {Euclidean Distance}[http://en.wikipedia.org/wiki/Euclidean_distance] or {Pearson's Correlation Coefficient}[http://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient] to calculate the similarity between persons and their preferences.
 
 == Installation:
 
@@ -18,15 +19,21 @@ At your Rails model that represents a person (can be an _User_, _Member_, or som
     # ...
   end
 
-  # ratings
+So, a person can rate things:
+
   current_user = User.create(:name => 'Diogenes')
   movie = Movie.create(:name => 'The Tourist', :year => 2010)
 
   current_user.rate!(movie, 4.5)
 
-  # after a lot of ratings... recommendations.
+And after a lot of ratings... *recommendations*:
+
   Movie.find_recommendations_for(current_user) # => movies and more movies...
 
+By default, the similarity strategy used is Euclidean Distance, but you can change that passing the _strategy_ option:
+  Movie.find_recommendations_for(current_user, :strategy => :pearson)
+
+
 == Contributing to coletivo
  
 * Check out the latest master to make sure the feature hasn't been implemented or the bug hasn't been fixed yet
@@ -41,4 +48,3 @@ At your Rails model that represents a person (can be an _User_, _Member_, or som
 
 Copyright (c) 2011 Diógenes Falcão. See LICENSE.txt for
 further details.
-

data/lib/coletivo/models/recommendable.rb

@@ -7,57 +7,15 @@ module Coletivo
       end
 
       module ClassMethods
-        def find_recommendations_for(model, options = {})
-          sim_totals, weighted_means = {}, {}
+        def find_recommendations_for(person, options = {})
           preferences = options[:preferences] ||=
-            load_preferences_for_recommendation(model)
+            load_preferences_for_recommendation(person)
+          top = predict_highest_ratings(person, preferences, options)
+          ids = top.collect(&:last)
 
-          preferences.each do |other, other_prefs|
-            next if other == model
-
-            sim = model.similarity_with(other, options)
-            next if sim <= 0
-
-            other_prefs.each do |item, weight|
-              unless preferences[model.id].keys.include?(item)
-                sim_totals[item] ||= 0
-                weighted_means[item] ||= 0
-
-                sim_totals[item] += sim
-                weighted_means[item] += weight * sim
-              end
-            end
-          end
-
-          # DESC sort by weighted mean of ratings
-          # e.g: [[5.35, "movie_2"], [2.0, "movie_4"]]
-          top = weighted_means.collect { |i, mean| [mean/sim_totals[i], i] }\
-            .sort { |t_one, t_other| t_other <=> t_one }
-
-          ids = top.collect(&:last) # e.g: ['movie_2', 'movie_4']
-          models = where(:id => ids)
-
-          top.collect { |weight, item| models.detect {|m| m.id == item } }\
-            .compact
+          where(:id => ids).limit(options[:limit]).all
         end
 
-        # Map a preferences Hash by a collection of ratings.
-        #
-        # Rating objects have a +person+, a +rateable+ object and a +weight+.
-        #
-        # The preferences can be mapped for _persons_, like that:
-        #   {
-        #      :person_1 => {:movie_1 => 2.0, :movie_2 => 5.0},
-        #      :person_2 => {:movie_1 => 3.0, :movie_2 => 4.0}
-        #   }
-        #
-        # Or for _rateable_ items, like that:
-        #   {
-        #      :movie_1 => {:person_1 => 2.0, :person_2 => 3.0},
-        #      :movie_2 => {:person_1 => 5.0, :person_2 => 4.0}
-        #   }
-        #
-        # Expected keys are :person or :rateable
         def map_ratings_to_preferences(ratings)
           #TODO: (???) Item based mapping.
           key, subkey = :person_id, :rateable_id
@@ -71,12 +29,40 @@ module Coletivo
           preferences
         end
 
-        def load_preferences_for_recommendation(model)
+        def load_preferences_for_recommendation(person)
           r = Coletivo::Config.ratings_container\
-                .find_for_recommendation(model, self)
+                .find_for_recommendation(person, self)
 
           map_ratings_to_preferences(r)
         end
+
+        private
+
+        def predict_highest_ratings(person, people_preferences, options)
+          data = {}
+          people_preferences.each do |other, other_prefs|
+            next if other == person
+
+            sim = person.similarity_with(other, options)
+            next if sim <= 0
+
+            other_prefs.each do |item, weight|
+              unless people_preferences[person.id].keys.include?(item)
+                data[item] ||= {:total_similarity => 0.0, :weighted_mean => 0.0}
+                data[item][:total_similarity] += sim
+                data[item][:weighted_mean] += weight * sim
+              end
+            end
+          end
+
+          # e.g: [[5.35, "movie_2"], [2.0, "movie_4"]]
+          guessed_rating_and_id = lambda { |item, item_data|
+            [item_data[:weighted_mean] / item_data[:total_similarity], item]
+          }
+
+          # DESC sorting by weighted mean of ratings
+          data.collect(&guessed_rating_and_id).sort_by(&:first).reverse
+        end
       end
 
       module InstanceMethods

data/lib/coletivo/similarity/euclidean_distance_strategy.rb

@@ -14,5 +14,5 @@ module Coletivo
       end
     end
   end
-end # Similarity
+end
 

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification 
 name: coletivo
 version: !ruby/object:Gem::Version 
-  hash: 29
+  hash: 27
   prerelease: false
   segments: 
   - 0
   - 0
-  - 1
-  version: 0.0.1
+  - 2
+  version: 0.0.2
 platform: ruby
 authors: 
 - "Di\xC3\xB3genes Falc\xC3\xA3o"
@@ -15,7 +15,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2011-07-20 00:00:00 -03:00
+date: 2011-09-20 00:00:00 -03:00
 default_executable: 
 dependencies: 
 - !ruby/object:Gem::Dependency