I have 3 classification rarity with probability something like this

class S has 10% probability

class A has 30% probability

class B has 60% probability

So i code like this

pool = ["S", "A", "A", "A", "B", "B", "B", "B", "B", "B"]
11.times do
 puts pool[rand(10) - 1]
end

and the result is quite correct in my guest (cmiiw)

B
A
B
B
S
B
A
S
B
A
B

but i become confuse when i should add more class and change S class probability to 1%

pool now become

class S has 1% probability

class A has 29% probability

class B has 30% probability

class C has 40% probability

i am not sure i should create pool like pool variable before because 1% is 1/10 is not integer.

Kindly need help, thanks!

CodePudding user response：

Consider using the aliastable gem. It’s a ruby implementation of the alias method published by A.J. Walker back in 1974.

The alias method uses O(n) time to set up the table with n outcomes, then O(1) time per value to generate. By contrast, setting up a CDF and then doing a binary search for values would take O(n) and O(log n) time respectively.

This little demo:

require 'aliastable'

values = %w(S A B C)
probs = [1r/100, 29r/100, 3r/10, 2r/5] # r => Rational

my_distribution = AliasTable.new(values, probs)
results = Array.new(1_000_000) { my_distribution.generate }
p results.tally

produces results such as

{"C"=>400361, "B"=>300121, "A"=>289636, "S"=>9882}

CodePudding user response：

pool = 100.times.map do
  r = rand

  if r <= 0.01
    'S'
  elsif r <= 0.30
    'A'
  elsif r <= 0.6
    'B'
  else
    'D'
  end
end
p pool
p pool.tally

This would output something like

["D", "B", "D", ....]
{"D"=>39, "B"=>28, "A"=>31, "S"=>2}

You could also force rand to return an Integer:

r = rand(0..100)

and then check for integers

if r <= 1

or use a case statement and check for ranges like in the in the answer linked by Stefan.

CodePudding user response：

Just for fun I adpated @Pascal's answer to a class with an Enumerator

class ProbabilityGenerator 
  include Enumerable
  attr_reader :probabilities

  def initialize(probabilities)
    raise ArgumentError unless probabilities.values.sum == 1.0
    @probabilities = prioritize(probabilities)
    @enum = Enumerator.new do |y| 
       v = rand
       loop do 
         y << @probabilities.find {|element, stat|  v <= stat}.first
         v = rand
       end 
    end 
  end 

  def each(&block)
    @enum.each(&block)
  end 

  def next 
    @enum.next
  end

  private 
    def prioritize(probabilities)
      probabilities
        .transform_values(&:to_r)
        .sort_by {|k,v| v}
        .tap {|a| a.each_cons(2) {|b,c| c[1] = b[1]   c[1]} } 
        .to_h
    end 
end 

You can use as follows:

pb = ProbabilityGenerator.new(s: 0.01, b: 0.6, a: 0.3, d: 0.09)
pb.take(10)
#=> [:a, :b, :d, :d, :b, :b, :b, :b, :d, :b]
# or just keep generating numbers 
pb.next 
#=> :a
pb.next 
#=> :b 

Running the same deterministics as @pjs

pb.take(1_000_000).tally
# => {:a=>300033, :b=>600362, :d=>89687, :s=>9918}