Class: Anomaly::Detector

Inherits:
Object
  • Object
show all
Defined in:
lib/anomaly/detector.rb

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(examples = nil, opts = {}) ⇒ Detector

Returns a new instance of Detector.



5
6
7
8
 
# File 'lib/anomaly/detector.rb', line 5

def initialize(examples = nil, opts = {})
  @m = 0
  train(examples, opts) if examples
end

Instance Attribute Details

#eps ⇒ Object

Returns the value of attribute eps.



3
4
5
 
# File 'lib/anomaly/detector.rb', line 3

def eps
  @eps
end

Instance Method Details

#anomaly?(x, eps = @eps) ⇒ Boolean

Returns:

  • (Boolean) 


76
77
78
 
# File 'lib/anomaly/detector.rb', line 76

def anomaly?(x, eps = @eps)
  probability(x) < eps
end

#probability(x) ⇒ Object

Limit the probability of features to [0,1] to keep probabilities at same scale.

Raises:

  • (ArgumentError) 


67
68
69
70
71
72
73
74
 
# File 'lib/anomaly/detector.rb', line 67

def probability(x)
  raise "Train me first" unless trained?
  raise ArgumentError, "First argument must have #{@n} elements" if x.size != @n
  @n.times.map do |i|
    p = normal_pdf(x[i], @mean[i], @std[i])
    (p.nan? or p > 1) ? 1 : p
  end.reduce(1, :*)
end

#train(examples, opts = {}) ⇒ Object 



10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
 
# File 'lib/anomaly/detector.rb', line 10

def train(examples, opts = {})
  raise "No examples" if examples.empty?
  raise "Must have at least two columns" if examples.first.size < 2

  # Divide into groups since we only want to train with non-anomalies.
  anomalies = []
  non_anomalies = []
  examples.each do |example|
    if example.last == 0
      non_anomalies << example
    else
      anomalies << example
    end
  end

  raise "Must have at least one non-anomaly" if non_anomalies.empty?

  @eps = (opts[:eps] || 0).to_f
  if @eps > 0
    # Use all non-anomalies to train.
    training_examples = non_anomalies
  else
    training_examples, test_examples = partition!(non_anomalies)
    test_examples.concat(anomalies)
  end
  # Remove last column.
  training_examples = training_examples.map{|e| e[0..-2]}
  @m = training_examples.size
  @n = training_examples.first.size

  if defined?(NMatrix)
    training_examples = NMatrix.to_na(training_examples)
    # Convert these to an Array for Marshal.dump
    @mean = training_examples.mean(1).to_a
    @std = training_examples.stddev(1).to_a
  else
    # Default to Array, since built-in Matrix does not give us a big performance advantage.
    cols = @n.times.map{|i| training_examples.map{|r| r[i]}}
    @mean = cols.map{|c| mean(c)}
    @std = cols.each_with_index.map{|c,i| std(c, @mean[i])}
  end
  @std.map!{|std| (std == 0 or std.nan?) ? Float::MIN : std}

  if @eps == 0
    # Find the best eps.
    epss = (1..9).map{|i| [1,3,5,7,9].map{|j| (j*10**(-i)).to_f }}.flatten
    f1_scores = epss.map{|eps| [eps, compute_f1_score(test_examples, eps)] }
    @eps, best_f1 = f1_scores.max_by{|v| v[1]}
  end
end

#trained? ⇒ Boolean

Returns:

  • (Boolean) 


61
62
63
 
# File 'lib/anomaly/detector.rb', line 61

def trained?
  @m > 0
end