data_mining 0.0.1

checksums.yaml

@@ -0,0 +1,7 @@
+---
+SHA1:
+  metadata.gz: 5da253c99db083a7e8a8694ef479299269b95c87
+  data.tar.gz: 873c011a2e26d9c1c11165a7ce9aa401d0d50250
+SHA512:
+  metadata.gz: 6f2f01babec563518f49ddd12ac23a55865245ab6b1c9862b88836d5544eb2be6ca49574f8261ed6f0d215e5880c14804dc5f54ae1757220b1bc41a71eeb780f
+  data.tar.gz: 027e465483ca27939f8b43167ae9aacfd3cd709545e9bfb38c3a5cc31c0d39d5f2e1b23ec74d4c3319e01c5b4552c2f9ec41f15140ae5cb34e4b6ea73185be32

data/lib/data_mining/dbscan.rb

@@ -0,0 +1,105 @@
+module DataMining
+  # Density-Based clustering / Outlier-Detection Algorithu
+  class DBScan
+    # Find clusters and outliers
+    #
+    # Example:
+    #   >> input      = [[:p1, 1], [:p2, 2], [:p3, 10]]
+    #   >> radius     = 3
+    #   >> min_points = 2
+    #   >> dbscan     = DataMining::DBScan.cluster(input, radius, min_points)
+    #   >> dbscan.build!
+    #   >>
+    #   >> dbscan.clusters # gives array of clusters found (:p1, :p2)
+    #   >>
+    #   >> dbscan.outliers # gives array of outliers found (:p3)
+    #
+    # Arguments:
+    #   data: (array of arrays, like [[:id, value], [:id2, value2]]
+    #   radius: (integer)
+    #   min_points: (integer)
+
+    def self.cluster(data, radius, min_points)
+      DBScan.new(data, radius, min_points)
+    end
+
+    def initialize(data, radius, min_points)
+      @data               = data.map { |i, v| DataMining::Point.new(i, v) }
+      @radius             = radius
+      @min_points         = min_points
+      @current_cluster_id = 0
+      @clusters           = {}
+      @unvisited_points   = @data.shuffle
+    end
+
+    def build!
+      dbscan
+      clusters
+    end
+
+    def outliers
+      @data.select { |p| !p.assigned_to_cluster? }
+    end
+
+    def clusters
+      @clusters.map { |cluster, points| { cluster => points.each(&:id) } }
+    end
+
+    private
+
+    def dbscan
+      until unvisited_points.empty?
+        p = unvisited_points.pop
+        p.visit!
+
+        neighborhood = get_neighborhood(p)
+        create_cluster(p, neighborhood) if core_object?(neighborhood)
+      end
+    end
+
+    def unvisited_points
+      @unvisited_points.select! { |p| !p.visited? }
+      @unvisited_points
+    end
+
+    def create_cluster(point, neighborhood)
+      @current_cluster_id += 1
+      point.assign_to_cluster!
+      (@clusters[@current_cluster_id] ||= []) << point
+      fill_current_cluster(neighborhood)
+    end
+
+    def fill_current_cluster(neighborhood)
+      neighborhood.each do |neighbor|
+        elaborate(neighbor) unless neighbor.visited?
+        neighbor.assign_to_cluster!
+      end
+    end
+
+    def elaborate(point)
+      point.visit!
+      @clusters[@current_cluster_id] << point unless point.assigned_to_cluster?
+      neighborhood = get_neighborhood(point)
+      fill_current_cluster(neighborhood) if core_object?(neighborhood)
+    end
+
+    # use map instead of each?
+    def get_neighborhood(point)
+      neighborhood = []
+      @data.each { |p| neighborhood << p if neighbors?(p, point) }
+      neighborhood
+    end
+
+    def core_object?(neighborhood)
+      return true if neighborhood.size >= (@min_points - 1)
+      false
+    end
+
+    def neighbors?(p1, p2)
+      return true if (p1.value - p2.value).abs <= @radius && p1 != p2
+      false
+    end
+  end
+end
+
+require 'data_mining/point'

data/lib/data_mining/point.rb

@@ -0,0 +1,35 @@
+module DataMining
+  # Point class
+  class Point
+    attr_reader :id, :value
+
+    # Represents a Point for the DBScan Algorithm
+    #
+    # Arguments:
+    #   id: (symbol)
+    #   value: (integer)
+
+    def initialize(id, value)
+      @id             = id
+      @value          = value
+      @visited        = false
+      @in_a_cluster   = false
+    end
+
+    def assigned_to_cluster?
+      @in_a_cluster
+    end
+
+    def assign_to_cluster!
+      @in_a_cluster = true
+    end
+
+    def visited?
+      @visited
+    end
+
+    def visit!
+      @visited = true
+    end
+  end
+end

data/lib/data_mining.rb

@@ -0,0 +1,4 @@
+module DataMining
+end
+
+require 'data_mining/dbscan'

metadata

@@ -0,0 +1,46 @@
+--- !ruby/object:Gem::Specification
+name: data_mining
+version: !ruby/object:Gem::Version
+  version: 0.0.1
+platform: ruby
+authors:
+- Manuel Stuefer
+autorequire: 
+bindir: bin
+cert_chain: []
+date: 2015-06-23 00:00:00.000000000 Z
+dependencies: []
+description: A collection of data mining algorithms
+email: mstuefer@gmail.com
+executables: []
+extensions: []
+extra_rdoc_files: []
+files:
+- lib/data_mining.rb
+- lib/data_mining/dbscan.rb
+- lib/data_mining/point.rb
+homepage: http://rubygems.org/gems/data_mining
+licenses:
+- MIT
+metadata: {}
+post_install_message: 
+rdoc_options: []
+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: 2.4.6
+signing_key: 
+specification_version: 4
+summary: Data-Mining-Algorithms
+test_files: []