decision-tree 0.0.2 → 0.0.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 6f70b8957300f4487b14eec4a2ba6461395899c8
-  data.tar.gz: 62425aa46fa2199efafee60cfcb13f860fa76c54
+  metadata.gz: 618f5b362f73985a29b4c14122ee3f8759819e37
+  data.tar.gz: 6ce2d965ce54258bce09533ca4504a547d3d45d2
 SHA512:
-  metadata.gz: e6c52758675d22c3d91493cdaaef808e791d53ad73097cbc582a869e68c977f4a8f5bef47c218ab581fed4f422c3e1605853597cf2e1c9fdeaceda6e5b65a7a2
-  data.tar.gz: a1bafd5b22b0d54215ef74edfc5fbd82bc2c3c75ce78543e1ada9367ce504e711aa0510487fe08e31e1f3e369ce3eb8b2c7f6b20d8c3a4a3c7eed35873f392a0
+  metadata.gz: cbe8ce578b1b98da49f7407cda609826d55216011d4634c0ea6782be63cd230a0eef4786b5497d2ca8aa71737bdc65a244308eaef348dcb28e0cd254c7e7b5ff
+  data.tar.gz: 06cc3773d1add266888dde7a37df1a5523342c4500811004e2d1a4f73617e53c6e817e3d4a4a05d5f670c67b20d535038c5cd0733da33ce273f6ae10657f6a79

data/lib/decision-tree.rb

@@ -74,20 +74,36 @@ module DecisionTree
 
 	        @path << choose_best_feature(entries)
 
-	        build_child_nodes(entries)
+	        if @algorithm == 'id3'
+	        	build_child_nodes(entries)
+	        elsif algorithm=='c45'
+	        	if feature_type=='num'
+	        		build_child_nodes_with_continuous_value(entries)
+	        	else
+	        		build_child_nodes(entries)
+	        	end
+	        end
 	    end
 
+
 		def feature_index
 			@path[-1]
 		end
 
 
 		def feature_name
-			@columns[ @path[-1] ]
+			@columns[ @path[-1] ].split(':')[0]
 		end
 
-	    def to_pseudo_code(buff=nil,indent="")
-	        buff = Array.new if buff.nil?
+
+		def feature_type
+			t = @columns[ @path[-1] ].split(':')[1]
+			t || 'string'
+		end
+
+
+		def to_pseudo_code(buff=nil,indent="")
+			buff = Array.new if buff.nil?
 
 	        if @child_nodes.size==0
 	            result = @labels.to_set.to_a
@@ -96,14 +112,27 @@ module DecisionTree
 	            else
 	                buff << "#{indent}return #{@labels}"
 	            end
+	            return buff
 	        end
 
-	        @child_nodes.each do |feature_value,child_node|
-	            buff << "#{indent}if(#{feature_name} == #{feature_value}){"
-	            # buff << "#{indent}if(#{feature_index} == #{feature_value}){"
-	            child_node.to_pseudo_code(buff, indent+"  " )
-	            buff << "#{indent}}"
+	        if @algorithm=='c45' and feature_type=='num'
+	        	sorted_nodes = @child_nodes.sort_by{|k,v| k.to_f }
+	        	sorted_nodes[1..-1].to_a.each do |feature_value,child_node|
+		            buff << "#{indent}if(#{feature_name} >= #{feature_value}){"
+		            child_node.to_pseudo_code(buff, indent+"  " )
+		            buff << "#{indent}}"
+		        end
+		        buff << "#{indent}else{"
+		        sorted_nodes[0][1].to_pseudo_code(buff, indent+"  " )
+		        buff << "#{indent}}"
+	        else
+	        	@child_nodes.each do |feature_value,child_node|
+		            buff << "#{indent}if(#{feature_name} == #{feature_value}){"
+		            child_node.to_pseudo_code(buff, indent+"  " )
+		            buff << "#{indent}}"
+		        end
 	        end
+
 	        return buff
 	    end
 
@@ -113,11 +142,24 @@ module DecisionTree
 	    		probability = Hash.new(0)
 	    		@labels.each{|k| probability[k] += 1 }
 	    		probability.each{|k,v| probability[k] = v / @labels.size.to_f }
-	    		return probability.to_json
+	    		return probability
 	    	else
-	    		feature_value = vector[feature_index]
-	    		return default if not @child_nodes.has_key?(feature_value)
-		    	return @child_nodes[feature_value].predict(vector)
+	    		if @algorithm=='c45' and feature_type=='num'
+	    			curr_value = vector[feature_index]
+
+	    			sorted_nodes = @child_nodes.sort_by{|k,v| k.to_f }
+	        		last_node = sorted_nodes[0][1]
+	        		sorted_nodes[1..-1].to_a.each do |feature_value,child_node|
+	        			break if curr_value.to_f < feature_value.to_f
+	        			last_node = child_node
+	        		end
+
+		    		return last_node.predict(vector,default)
+	    		else
+	    			feature_value = vector[feature_index]
+	    			return default if not @child_nodes.has_key?(feature_value)
+		    		return @child_nodes[feature_value].predict(vector,default)
+		    	end
 		    end
 	    end
 
@@ -130,21 +172,16 @@ module DecisionTree
 			@dimension.times do |i|
 				next if @path.include?(i)
 				child_entropy = entries.map{|x| x[:features][i]}.concitional_entropy_with(labels)
-				
+
 	            ig = if @algorithm=='id3'
 	                @entropy - child_entropy
 	            else# c45
-	                a = (@entropy - child_entropy)
-	                b = entries.map{|x| x[:features][i]}.entropy
-	                # puts "@path=#{@path}"
-	                # puts "i=#{i} @entropy=#{@entropy} child_entropy=#{child_entropy} a=#{a} b=#{b}"
 	                gain = (@entropy - child_entropy) / entries.map{|x| x[:features][i]}.entropy
 	                gain = 0 if gain.nan?
 	                gain
 	            end
 
 	            max_ig = {index: i, ig: ig} if ig > max_ig[:ig]
-	            # puts "max_ig=#{max_ig} ig=#{ig}"
 			end
 			return max_ig[:index]
 		end
@@ -162,5 +199,32 @@ module DecisionTree
 				@child_nodes[feature_value] = Node.new(child_entries, @columns, @algorithm, @dimension, self, feature_value, @path.dup)
 			end
 		end
+
+
+		def build_child_nodes_with_continuous_value(entries)
+
+			buff = Hash.new{|h,feature_value| h[feature_value] = Array.new}
+			sorted_entries = entries.sort_by{|e| e[:features][feature_index].to_f }
+
+			last_label = nil #sorted_entries[0][:label].to_s
+			last_value = nil # sorted_entries[0][:features][feature_index]
+
+			sorted_entries.each_with_index do |e, i|
+
+				feature_value = e[:features][feature_index]
+
+				if last_label != e[:label].to_s
+					last_value = feature_value.to_s
+					last_label = e[:label].to_s
+				end
+
+				buff[last_value] << e
+			end
+
+			buff.each do |feature_value,child_entries|
+				@child_nodes[feature_value] = Node.new(child_entries, @columns, @algorithm, @dimension, self, feature_value, @path.dup)
+			end
+
+		end
 	end
-end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: decision-tree
 version: !ruby/object:Gem::Version
-  version: 0.0.2
+  version: 0.0.3
 platform: ruby
 authors:
 - ireullin
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2017-04-17 00:00:00.000000000 Z
+date: 2017-04-19 00:00:00.000000000 Z
 dependencies: []
 description: 'A decision tree library which implemented ID3 & C4.5 of algorithms '
 email: