lib/svm/svm.rb

   1 require 'gnuplot'
   2
   3 class ClassifiedData
   4   attr_accessor :data, :classification
   5
   6   def initialize(data = Array.new, classification = nil)
   7     @data = data
   8     @classification = classification
   9   end
  10 end
  11
  12 def load_data(filename, only_numbers = false)
  13   rows = []
  14   IO.foreach(filename) do |line|
  15     rows << ClassifiedData.new(line.chomp.split(',')[0..-2].map {|field| only_numbers ? field.to_f : field},
  16                                line.chomp.split(',')[-1].to_i)
  17   end
  18   rows
  19 end
  20
  21
  22 def plot_age_matches(rows)
  23   Gnuplot.open do |gp|
  24     Gnuplot::Plot.new( gp ) do |plot|
  25
  26       plot.title  "Ages of matches"
  27       plot.ylabel "man"
  28       plot.xlabel "woman"
  29
  30       matches = rows.select {|r| r.classification == 1}.map {|r| r.data}
  31       non_matches = rows.select {|r| r.classification == 0}.map {|r| r.data}
  32
  33       plot.data = [ Gnuplot::DataSet.new( [matches] ) do |ds|
  34         ds.with = "points"
  35         ds.notitle
  36       end , Gnuplot::DataSet.new( [non_matches] ) do |ds|
  37         ds.with = "points"
  38         ds.notitle
  39       end ]
  40     end
  41   end
  42 end
  43
  44
  45 def linear_train(rows)
  46   sums = {}
  47   averages = {}
  48   counts = Hash.new(0)
  49
  50   rows.each do |row|
  51     row_class = row.classification
  52     sums[row_class] ||= [0.0] * (row.data.length)
  53
  54     (row.data.length).times do |i|
  55       sums[row_class][i] += row.data[i]
  56     end
  57     counts[row_class] += 1
  58   end
  59
  60   sums.keys.each do |match_class|
  61     averages[match_class] = sums[match_class].map {|sum| sum / counts[match_class]}
  62   end
  63   averages
  64 end
  65
  66 def dot_product(v1, v2)
  67   (v1.zip v2).map {|c| c.reduce(:*) }.reduce(:+)
  68 end
  69
  70 def dot_product_classify(point, averages)
  71   b = (dot_product(averages[1], averages[1]) - dot_product(averages[0], averages[0])) / 2
  72   y = dot_product(point, averages[0]) - dot_product(point, averages[1]) + b
  73   if y > 0
  74     0
  75   else
  76     1
  77   end
  78 end
  79
  80 def yesno(v)
  81   if v == 'yes' then 1
  82   elsif v == 'no' then -1
  83   else 0
  84   end
  85 end
  86
  87 def match_count(interests1, interests2)
  88   (interests1.split(':') & interests2.split(':')).length
  89 end
  90
  91 def miles_distance(a1, a2)
  92   0
  93 end
  94
  95