Module: ReplaceMissingValues

Included in:
FSelector::Base
Defined in:
lib/fselector/replace_missing_values.rb

Overview

replace missing feature values

Instance Method Summary collapse

Instance Method Details

#replace_by_fixed_value!(val)

Note:

data structure will be altered

replace missing feature value by a fixed value, applicable to both discrete and continuous feature



11
12
13
14
15
16
17
18
19
20
21
22
 
# File 'lib/fselector/replace_missing_values.rb', line 11

def replace_by_fixed_value!(val)
  each_sample do |k, s|
    each_feature do |f|
      if not s.has_key? f
        s[f] = val
      end
    end
  end
  
  # clear variables
  clear_vars
end

#replace_by_knn_value!(k = 1)

Note:

data structure will be altered, and the nearest neighbors are determined by Euclidean distance

replace missing feature value by weighted k-nearest neighbors' value, applicable only to continuous feature

val = sigma_k (val_k * w_k)

where w_k = (sum_d - d_k) / ((K-1) * sum_d)
      sum_d = sigma_k (d_k)
      K: number of d_k
      sigma_k (w_k) = 1, normalized to 1

ref: Microarray missing data imputation based on a set theoretic framework and biological knowledge

Parameters:

  • k (Integer) (defaults to: 1) —

    number of nearest neighbors



134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
 
# File 'lib/fselector/replace_missing_values.rb', line 134

def replace_by_knn_value!(k=1)
  each_sample do |ki, si|
    # potential features having missing value
    mv_fs = get_features - si.keys
    next if mv_fs.empty? # sample si has no missing value
    
    # record object value for each feature missing value
    f2val = {}
    mv_fs.each do |mv_f|
      knn_s, knn_d = [], []
      
      each_sample do |kj, sj|
        # sample sj also has missing value of mv_f
        next if not sj.has_key? mv_f
        
        d = euclidean_distance(si, sj)
        idx = knn_d.index { |di| d<di }
        
        if idx
          knn_s.insert(idx, sj)
          knn_d.insert(idx, d)
          
          if knn_s.size > k
            knn_s = knn_s[0...k]
            knn_d = knn_d[0...k]
          end
        else
          if knn_s.size < k
            knn_s << sj
            knn_d << d
          end
        end
      end
      
      # distance-weighted value from knn
      knn_d_sum = knn_d.sum
      sz = knn_d.size
      val = 0.0
      knn_s.each_with_index do |s, i|
        if sz > 1
          if not knn_d_sum.zero?
            val += s[mv_f] * (knn_d_sum-knn_d[i]) / ((sz-1)*knn_d_sum)
          else
            val += s[mv_f] * 1.0 / sz
          end
        else # only one nearest neighbor
          val = s[mv_f]
        end
      end
      
      f2val[mv_f] = val
     #pp [si, mv_f, knn_s, knn_d, val]
    end
    
    # set value
    f2val.each do |f, v|
      si[f] = v
    end
  end
  
  # clear variables
  clear_vars
end

#replace_by_mean_value!(mode = :by_column)

Note:

data structure will be altered

replace missing feature value by mean feature value, applicable only to continuous feature

Parameters:

  • mode (Symbol) (defaults to: :by_column) —

    column or row mode

    • :by_column # use the mean value of the same feature among all instances
    • :by_row # use the mean value of different features in current instance


35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
 
# File 'lib/fselector/replace_missing_values.rb', line 35

def replace_by_mean_value!(mode = :by_column)
  each_sample do |k, s|
    mean = s.values.mean if mode == :by_row
    
    each_feature do |f|
      fv = get_feature_values(f)
      next if fv.size == get_sample_size # no missing values
      
      mean = fv.ave if mode == :by_column
      if not s.has_key? f
        s[f] = mean
      end
    end
  end
  
  # clear variables
  clear_vars
end

#replace_by_median_value!(mode = :by_column)

Note:

data structure will be altered

replace missing feature value by median feature value, applicable only to continuous feature

Parameters:

  • mode (Symbol) (defaults to: :by_column) —

    column or row mode

    • :by_column # use the mean value of the same feature among all instances
    • :by_row # use the mean value of different features in current instance


65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
 
# File 'lib/fselector/replace_missing_values.rb', line 65

def replace_by_median_value!(mode = :by_column)
  each_sample do |k, s|
    median = s.values.median if mode == :by_row
    
    each_feature do |f|
      fv = get_feature_values(f)
      next if fv.size == get_sample_size # no missing values
      
      median = fv.median if mode == :by_column
      if not s.has_key? f
        s[f] = median
      end
    end
  end
  
  # clear variables
  clear_vars
end

#replace_by_most_seen_value!

Note:

data structure will be altered

replace missing feature value by most seen feature value, applicable only to discrete feature



91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
 
# File 'lib/fselector/replace_missing_values.rb', line 91

def replace_by_most_seen_value!
  each_sample do |k, s|
    each_feature do |f|
      fv = get_feature_values(f)
      next if fv.size == get_sample_size # no missing values
      
      seen_count, seen_value = 0, nil
      fv.uniq.each do |v|
        count = fv.count(v)
        if count > seen_count
          seen_count = count
          seen_value = v
        end
      end
      
      if not s.has_key? f
        s[f] = seen_value
      end
    end
  end
  
  # clear variables
  clear_vars
end