agglomerative_clustering 0.0.3 → 0.0.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 03755a284a11c8225365f9b22df38da433e07994
-  data.tar.gz: 6996369a3b16c5541cd694a8315c2c40606b91e3
+  metadata.gz: 3c8adc349d03e25a7a982a0479069f510eb6a97a
+  data.tar.gz: 91cfe58dd5b19999699d1cf331bcd12b0b2451db
 SHA512:
-  metadata.gz: 1061fb7aae2bc9c6cee7ba054593331a6e9db4c5388d895f821aad46b6d05d86721f61c947ce5377095ff7dc3adf8f93494ce9a3697f44357521d5ecc97e8523
-  data.tar.gz: f481ee1c97df283e3a354ed447319fbf9041d5507ac677269288e9c7802eb743510e016a6a0fca9a5232eb0a98420a1c393e1a08c9a52a0dfac91fcc21d91fd7
+  metadata.gz: 7ecefeb4ecb8f02f10da56b95cb883c6752631a28fc0212fb8e8f72f3b57d361c7597179ba9bf97c4a804f0540f7a81393a0c0263048994c3b14bca63bc6e364
+  data.tar.gz: 8b979ffcc0778884639e02ae3651792582802efd31e58ccd8d2964d1d56b30b94dd00953389a3df55654287412d4bd8ef64321a65ab3af3386570106fdce6969

data/cluster_example.rb

@@ -0,0 +1,93 @@
+#!/usr/bin/env ruby
+require 'agglomerative_clustering'
+
+puts "Enter a total number of points"
+points = gets.chomp.to_i
+puts "Please enter a P value (Percentage of points to determine outliers)"
+percentage = gets.chomp.to_f
+puts "Please enter a d value (Distance of points to determine outliers)"
+distance = gets.chomp.to_f
+puts "Please enter a number of clusters"
+k = gets.chomp.to_i
+puts "Enter y to enter a center point for each pseudo cluster (Or press enter for random)"
+center_points = []
+cp_input = gets.chomp
+if cp_input == 'y'
+  for l in 0..k-1
+    center_points << gets.chomp.split(',').map(&:to_i)
+  end
+else
+  for i in 0..k-1
+    center_points << [Random.rand(-100..100), Random.rand(-100..100), Random.rand(-100..100)]
+  end
+end
+
+puts "Processing.."
+min_set = AgglomerativeClustering::Set.new(AgglomerativeClustering::Linkage::Single.new)
+max_set = AgglomerativeClustering::Set.new(AgglomerativeClustering::Linkage::Complete.new)
+average_set = AgglomerativeClustering::Set.new(AgglomerativeClustering::Linkage::Average.new)
+center_set = AgglomerativeClustering::Set.new(AgglomerativeClustering::Linkage::Center.new)
+sets = [min_set, max_set, average_set, center_set]
+
+# Create set of 500 points based on provided parameters
+radius = 25
+for clusters in 0..k-1
+  cp = center_points[clusters]
+  for i in 0..(points/k-1)
+    x = cp[0] + (Random.rand(-radius..radius))
+    y = cp[1] + (Random.rand(-radius..radius))
+    z = cp[2] + (Random.rand(-radius..radius))
+    p = Point.new(x,y,z)
+    sets.each {|set| set.push(p)}
+  end
+end
+
+# Find and Remove Outliers in the Set
+sets.each do |set|
+  set.find_outliers(percentage, distance)
+end
+
+# Output outliers to the console
+if min_set.outliers.any?
+  puts 'Outliers Removed from Set:'
+  min_set.outliers.each do |outlier|
+    puts outlier
+  end
+else
+  puts "There are no outliers where #{percentage}% of the points lie at a distance greater than #{distance}"
+end
+
+# Run the clustering algorithms
+min_clusters = min_set.cluster(k)
+max_clusters = max_set.cluster(k)
+average_clusters = average_set.cluster(k)
+center_clusters = center_set.cluster(k)
+
+# Outputs points on each cluster
+puts "=====Minimum Linkage Clusters====="
+min_clusters.each_with_index{|c, i| puts "Cluster #{i + 1}:\n #{c.points}" }
+puts "=====Maximum Linkage Clusters====="
+max_clusters.each_with_index{|c, i| puts "Cluster #{i + 1}:\n #{c.points}" }
+puts "=====Average Linkage Clusters====="
+average_clusters.each_with_index{|c, i| puts "Cluster #{i + 1}:\n #{c.points}" }
+puts "=====Center Linkage Clusters======"
+center_clusters.each_with_index{|c, i| puts "Cluster #{i + 1}:\n #{c.points}" }
+
+# Output Silhouette Coefficients for the clusters in each set
+puts 'Minimum Linkage Silhouette Coefficients: '
+s1 = min_clusters.map { |c| AgglomerativeClustering::SilhouetteCoefficient.new(c).measure(min_clusters) }
+p s1
+puts 'Maximum Linkage Silhouette Coefficients: '
+s2 = max_clusters.map { |c| AgglomerativeClustering::SilhouetteCoefficient.new(c).measure(max_clusters) }
+p s2
+puts 'Average Linkage Silhouette Coefficients: '
+s3 = average_clusters.map { |cluster| AgglomerativeClustering::SilhouetteCoefficient.new(cluster).measure(average_clusters) }
+p s3
+puts 'Center Linkage Silhouette Coefficients: '
+s4 = center_clusters.map { |cluster| AgglomerativeClustering::SilhouetteCoefficient.new(cluster).measure(center_clusters) }
+p s4
+
+# output best average silhouette coefficient
+type_hash = {0 => 'Minimum', 1 => 'Maximum', 2 => 'Average', 3 => 'Center'}
+all = [s1,s2,s3,s4].map { |s| s.inject(:+)/s.size }
+puts "The best Average Silhouette Coefficient is #{all.max} from #{type_hash[all.index(all.max)]} Linkage"

data/lib/agglomerative_clustering/cluster.rb

@@ -8,6 +8,8 @@ module AgglomerativeClustering
 
     def points
       @points ||= []
+      @points.sort_by!{|p| [p.x, p.y, p.z] } if @points.any?
+      @points
     end
 
     def merge(cluster)

data/lib/agglomerative_clustering/linkage/center.rb

@@ -8,9 +8,10 @@ module AgglomerativeClustering
       end
 
       def center_point cluster
-        cluster.points.first.zip(*cluster.points[1..cluster.points.size-1]).map { |a,b| (a + b)/cluster.points.size.to_f }
+        return cluster.points.first if cluster.points.size == 1
+        cluster.points.first.zip(*cluster.points[1..cluster.points.size-1]).map { |a| a.inject(:+).to_f/a.size.to_f }
       end
-      
+
     end
   end
 end

data/lib/agglomerative_clustering/set.rb

@@ -38,9 +38,12 @@ module AgglomerativeClustering
       distance_matrix.matrix.each_with_index do |index, row, column|
         count_hash[row] ||= 0
         count_hash[row] += 1 if distance_matrix.matrix[row, column] > distance
-        set_outliers << points[row] if count_hash[row]/(distance_matrix.matrix.row_count - 1) > percentage_of_clusters/100
+        if count_hash[row]/(distance_matrix.matrix.row_count - 1) > percentage_of_clusters/100
+          set_outliers << points[row]
+        end
       end
       points.reject! { |point| outliers.include?(point) }
+      @distance_matrix = build_distance_matrix
       outliers
     end
 

data/lib/agglomerative_clustering/silhouette_coefficient.rb

@@ -7,6 +7,8 @@ module AgglomerativeClustering
       @main_cluster = main_cluster
     end
 
+    # Measures the silhouette coefficient of a cluster compared to all other clusters
+    # Returns the average silhouette coefficient of a cluster
     def measure clusters
       silhouettes = []
       average_distances = []
@@ -15,23 +17,25 @@ module AgglomerativeClustering
           (clusters - [main_cluster]).each do |cluster|
             distances = []
             cluster.points.each do |point2|
-              distances << euclidean_distance(point1, point2)
+              distances << euclidean_distance(point1, point2).round(2)
             end
             average_distances << distances.inject(:+)/distances.size
           end
-          b1 = average_distances.min
+          b1 = average_distances.min || 0
           s1 = (b1 - a1)/[a1,b1].max
           silhouettes << s1
         end
-      silhouettes.inject(:+) / silhouettes.size
+      (silhouettes.inject(:+) / silhouettes.size).round(2)
     end
 
+    # Calculates the a1 value of a cluster
     def calculate_a1 point
       distances = []
       main_cluster.points.each do |point1|
-        distances << euclidean_distance(point, point1)
+        distances << euclidean_distance(point, point1).round(2)
       end
-      distances.inject(:+)/(distances.size - 1)
+      return 0 if distances.size == 1
+      (distances.inject(:+)/(distances.size - 1)).round(2)
     end
 
   end

data/lib/agglomerative_clustering/version.rb

@@ -1,3 +1,3 @@
 module AgglomerativeClustering
-  VERSION = "0.0.3"
+  VERSION = "0.0.4"
 end

data/spec/lib/agglomerative_clustering/cluster_spec.rb

@@ -5,7 +5,7 @@ describe AgglomerativeClustering::Cluster do
       cluster1 = FactoryGirl.build(:cluster)
       cluster2 = FactoryGirl.build(:cluster)
       points = cluster1.points + cluster2.points
-      expect(cluster1.merge(cluster2).points).to eql(points)
+      expect(cluster1.merge(cluster2).points).to eql(points.sort_by{|p| [p.x, p.y, p.z]})
     end
   end
 end

data/spec/lib/agglomerative_clustering/linkage/single_spec.rb

@@ -3,17 +3,17 @@ describe AgglomerativeClustering::Linkage::Single do
   context '#calculate_distance' do
     it 'will calculate distance between clusters based on the min distnace between points' do
       single_linkage = AgglomerativeClustering::Linkage::Single.new
-      min_point = FactoryGirl.build(:point, x: 2, y: 2, z: 2)
-      max_point = FactoryGirl.build(:point, x: 5, y: 5, z: 5)
-      cluster1 = AgglomerativeClustering::Cluster.new(min_point)
+      min_point1 = FactoryGirl.build(:point, x: 2, y: 2, z: 2)
+      min_point2 = FactoryGirl.build(:point, x: 5, y: 5, z: 5)
+      cluster1 = AgglomerativeClustering::Cluster.new(min_point1)
       cluster2 = AgglomerativeClustering::Cluster.new(FactoryGirl.build(:point, x: 1, y: 1, z: 1))
       cluster3 = AgglomerativeClustering::Cluster.new(FactoryGirl.build(:point, x: 7, y: 7, z: 7))
-      cluster4 = AgglomerativeClustering::Cluster.new(max_point)
+      cluster4 = AgglomerativeClustering::Cluster.new(min_point2)
       cluster1 = cluster1.merge(cluster2)
       cluster3 = cluster3.merge(cluster4)
-      min_distance = single_linkage.euclidean_distance(min_point, max_point)
+      min_distance = single_linkage.euclidean_distance(min_point1, min_point2)
       expect(single_linkage.calculate_distance(cluster1, cluster3)).to eql(min_distance)
     end
   end
-  
+
 end

data/spec/lib/agglomerative_clustering/silhouette_coefficient_spec.rb

@@ -18,7 +18,7 @@ describe AgglomerativeClustering::SilhouetteCoefficient do
       cluster4.merge(cluster5).merge(cluster6)
       sc = AgglomerativeClustering::SilhouetteCoefficient.new(cluster1)
       clusters = [cluster1, cluster4]
-      expect(sc.measure(clusters).round(4)).to eql(-0.0893)
+      expect(sc.measure(clusters)).to eql(-0.09)
     end
   end
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: agglomerative_clustering
 version: !ruby/object:Gem::Version
-  version: 0.0.3
+  version: 0.0.4
 platform: ruby
 authors:
 - Bryan Mulvihill
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2014-11-02 00:00:00.000000000 Z
+date: 2014-11-15 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -67,7 +67,7 @@ files:
 - README.md
 - Rakefile
 - agglomerative_clustering.gemspec
-- cluster.rb
+- cluster_example.rb
 - lib/agglomerative_clustering.rb
 - lib/agglomerative_clustering/cluster.rb
 - lib/agglomerative_clustering/distance_matrix.rb

data/cluster.rb

@@ -1,62 +0,0 @@
-#!/usr/bin/env ruby
-require 'agglomerative_clustering'
-
-set = AgglomerativeClustering::Set.new(AgglomerativeClustering::Linkage::Single.new)
-for i in 0..99
-  x = Random.rand(0..100)
-  y = Random.rand(0..100)
-  z = Random.rand(0..100)
-  p = Point.new(x,y,z)
-  set.push(p)
-end
-for i in 100..199
-  x = Random.rand(200..299)
-  y = Random.rand(200..299)
-  z = Random.rand(200..299)
-  p = Point.new(x,y,z)
-  set.push(p)
-end
-for i in 200..299
-  x = Random.rand(400..499)
-  y = Random.rand(400..499)
-  z = Random.rand(400..499)
-  p = Point.new(x,y,z)
-  set.push(p)
-end
-
-percentage = 80
-distance = 150
-
-open('points.csv', 'w') do |f|
-  set.points.each do |point|
-    f << "#{point.x},#{point.y},#{point.z}\n"
-  end
-end
-
-open('outliers.csv', 'w') do |f|
-  set.find_outliers(percentage, distance).each do |point|
-    f << "#{point.x},#{point.y},#{point.z}\n"
-  end
-end
-
-if set.outliers.any?
-  puts 'Outliers Removed from Set:'
-  set.outliers.each do |outlier|
-    puts outlier
-  end
-else
-  puts "There are no outliers where #{percentage}% of the points lie at a distance greater than #{distance}"
-end
-
-clusters = set.cluster(3)
-clusters.each_with_index do |cluster, index|
-  open("cluster#{index}.csv", 'w') do |f|
-    cluster.points.each do |point|
-      f << "#{point.x},#{point.y},#{point.z}\n"
-    end
-  end
-end
-
-puts 'Silhouette Coefficient of First Cluster: '
-sc = AgglomerativeClustering::SilhouetteCoefficient.new(clusters[0])
-puts sc.measure(clusters)