1
2
3
4
5
6
7
8
9
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
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
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
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
|
require "lib/prediction_data.rb"
module Lib
module Util
def self.compute_variance( old_variance, n, new_mean, old_mean, new_value )
# use revursiv formular for computing the variance
# ( see Tysiak, Folgen: explizit und rekursiv, ISSN: 0025-5866
# http://www.frl.de/tysiakpapers/07_TY_Papers.pdf )
return (n>1 ? old_variance * (n-2)/(n-1) : 0) +
(new_mean - old_mean)**2 +
(n>1 ? (new_value - new_mean)**2/(n-1) : 0 )
end
end
class Predictions
def identifier(instance_index)
return instance_index.to_s
end
def initialize( data )
raise unless data.is_a?(Hash)
@feature_type = data[:feature_type]
@accept_values = data[:accept_values]
@num_classes = 1
#puts "predicted: "+predicted_values.inspect
#puts "actual: "+actual_values.inspect
#puts "confidence: "+confidence_values.inspect
raise "unknown feature_type: '"+@feature_type.to_s+"'" unless
@feature_type=="classification" || @feature_type=="regression"
raise "no predictions" if data[:predicted_values].size == 0
num_info = "predicted:"+data[:predicted_values].size.to_s+
" confidence:"+data[:confidence_values].size.to_s+" actual:"+data[:actual_values].size.to_s
raise "illegal num actual values "+num_info if data[:actual_values].size != data[:predicted_values].size
raise "illegal num confidence values "+num_info if data[:confidence_values].size != data[:predicted_values].size
case @feature_type
when "classification"
raise "accept_values missing while performing classification" unless @accept_values
@num_classes = @accept_values.size
raise "num classes < 2" if @num_classes<2
when "regression"
raise "accept_values != nil while performing regression" if @accept_values
end
@predicted_values = []
@actual_values = []
@confidence_values = []
init_stats()
(0..data[:predicted_values].size-1).each do |i|
update_stats( data[:predicted_values][i], data[:actual_values][i], data[:confidence_values][i] )
end
end
def init_stats
@conf_provided = false
@num_no_actual_value = 0
@num_with_actual_value = 0
@num_predicted = 0
@num_unpredicted = 0
@mean_confidence = 0
case @feature_type
when "classification"
# confusion-matrix will contain counts for predictions in a 2d array:
# index of first dim: actual values
# index of second dim: predicited values
# example:
# * summing up over all i with fixed n
# * confusion_matrix[i][n]
# * will give the number of instances that are predicted as n
@confusion_matrix = []
@accept_values.each do |v|
@confusion_matrix.push( Array.new( @num_classes, 0 ) )
end
@num_correct = 0
@num_incorrect = 0
when "regression"
@sum_error = 0
@sum_abs_error = 0
@sum_squared_error = 0
@prediction_mean = 0
@actual_mean = 0
@variance_predicted = 0
@variance_actual = 0
@sum_actual = 0
@sum_predicted = 0
@sum_multiply = 0
@sum_squares_actual = 0
@sum_squares_predicted = 0
@sum_confidence = 0
@weighted_sum_actual = 0
@weighted_sum_predicted = 0
@weighted_sum_multiply = 0
@weighted_sum_squares_actual = 0
@weighted_sum_squares_predicted = 0
@sum_weighted_abs_error = 0
@sum_weighted_squared_error = 0
end
end
def update_stats( predicted_value, actual_value, confidence_value )
raise "illegal confidence value: '"+confidence_value.to_s+"'" unless
confidence_value==nil or (confidence_value.is_a?(Numeric) and confidence_value>=0 and confidence_value<=1)
case @feature_type
when "classification"
{ "predicted"=>predicted_value, "actual"=>actual_value }.each do |s,v|
raise "illegal "+s+" classification-value ("+v.to_s+"),"+
"has to be either nil or index of predicted-values" if v!=nil and (!v.is_a?(Numeric) or v<0 or v>@num_classes)
end
when "regression"
{ "predicted"=>predicted_value, "actual"=>actual_value }.each do |s,v|
raise "illegal "+s+" regression-value ("+v.to_s+"),"+
" has to be either nil or number (not NaN, not Infinite)" unless v==nil or (v.is_a?(Numeric) and !v.nan? and v.finite?)
end
end
@predicted_values << predicted_value
@actual_values << actual_value
@confidence_values << confidence_value
if actual_value==nil
@num_no_actual_value += 1
else
@num_with_actual_value += 1
if predicted_value==nil
@num_unpredicted += 1
else
@num_predicted += 1
@conf_provided |= confidence_value!=nil
@mean_confidence = (confidence_value + @mean_confidence*(@num_predicted-1)) / @num_predicted.to_f if @conf_provided
case @feature_type
when "classification"
@confusion_matrix[actual_value][predicted_value] += 1
if (predicted_value == actual_value)
@num_correct += 1
else
@num_incorrect += 1
end
when "regression"
delta = predicted_value - actual_value
@sum_error += delta
@sum_abs_error += delta.abs
@sum_weighted_abs_error += delta.abs*confidence_value if @conf_provided
@sum_squared_error += delta**2
@sum_weighted_squared_error += (delta**2)*confidence_value if @conf_provided
old_prediction_mean = @prediction_mean
@prediction_mean = (@prediction_mean * (@num_predicted-1) + predicted_value) / @num_predicted.to_f
old_actual_mean = @actual_mean
@actual_mean = (@actual_mean * (@num_predicted-1) + actual_value) / @num_predicted.to_f
@variance_predicted = Util.compute_variance( @variance_predicted, @num_predicted,
@prediction_mean, old_prediction_mean, predicted_value )
@variance_actual = Util.compute_variance( @variance_actual, @num_predicted,
@actual_mean, old_actual_mean, actual_value )
@sum_actual += actual_value
@sum_predicted += predicted_value
@sum_multiply += (actual_value*predicted_value)
@sum_squares_actual += actual_value**2
@sum_squares_predicted += predicted_value**2
if @conf_provided
w_a = actual_value * confidence_value
w_p = predicted_value * confidence_value
@weighted_sum_actual += w_a
@weighted_sum_predicted += w_p
@weighted_sum_multiply += (w_a*w_p) if @conf_provided
@weighted_sum_squares_actual += w_a**2 if @conf_provided
@weighted_sum_squares_predicted += w_p**2 if @conf_provided
end
end
end
end
end
def percent_correct
raise "no classification" unless @feature_type=="classification"
pct = 100 * @num_correct / (@num_with_actual_value - @num_unpredicted).to_f
pct.nan? ? 0 : pct
end
def percent_incorrect
raise "no classification" unless @feature_type=="classification"
return 0 if @num_with_actual_value==0
return 100 * @num_incorrect / (@num_with_actual_value - @num_unpredicted).to_f
end
def accuracy
acc = percent_correct / 100.0
acc.nan? ? 0 : acc
end
def weighted_accuracy
return 0 unless confidence_values_available?
raise "no classification" unless @feature_type=="classification"
total = 0
correct = 0
(0..@predicted_values.size-1).each do |i|
if @predicted_values[i]!=nil
total += @confidence_values[i]
correct += @confidence_values[i] if @actual_values[i]==@predicted_values[i]
end
end
if total==0 || correct == 0
return 0
else
return correct / total
end
end
def percent_unpredicted
return 0 if @num_with_actual_value==0
return 100 * @num_unpredicted / @num_with_actual_value.to_f
end
def num_unpredicted
@num_unpredicted
end
def percent_without_class
return 0 if @predicted_values==0
return 100 * @num_no_actual_value / @predicted_values.size.to_f
end
def num_without_class
@num_no_actual_value
end
def num_correct
raise "no classification" unless @feature_type=="classification"
return @num_correct
end
def num_incorrect
raise "no classification" unless @feature_type=="classification"
return @num_incorrect
end
def num_unclassified
raise "no classification" unless @feature_type=="classification"
return @num_unpredicted
end
# internal structure of confusion matrix:
# hash with keys: hash{ :confusion_matrix_actual => <class_value>, :confusion_matrix_predicted => <class_value> }
# and values: <int-value>
def confusion_matrix
raise "no classification" unless @feature_type=="classification"
res = {}
(0..@num_classes-1).each do |actual|
(0..@num_classes-1).each do |predicted|
res[{:confusion_matrix_actual => @accept_values[actual],
:confusion_matrix_predicted => @accept_values[predicted]}] = @confusion_matrix[actual][predicted]
end
end
return res
end
# returns acutal values for a certain prediction
def confusion_matrix_row(predicted_class_index)
r = []
(0..@num_classes-1).each do |actual|
r << @confusion_matrix[actual][predicted_class_index]
end
return r
end
def area_under_roc(class_index=nil)
return prediction_feature_value_map( lambda{ |i| area_under_roc(i) } ) if
class_index==nil
return 0 unless confidence_values_available?
LOGGER.warn("TODO: implement approx computiation of AUC,"+
"so far Wilcoxon-Man-Whitney is used (exponential)") if
@predicted_values.size>1000
#puts "COMPUTING AUC "+class_index.to_s
tp_conf = []
fp_conf = []
(0..@predicted_values.size-1).each do |i|
if @predicted_values[i]!=nil
c = @confidence_values[i] * (@predicted_values[i]==class_index ? 1 : -1)
if @actual_values[i]==class_index
tp_conf << c
else
fp_conf << c
end
end
end
#puts tp_conf.inspect+"\n"+fp_conf.inspect+"\n\n"
return 0.0 if tp_conf.size == 0
return 1.0 if fp_conf.size == 0
sum = 0
tp_conf.each do |tp|
fp_conf.each do |fp|
sum += 1 if tp>fp
sum += 0.5 if tp==fp
end
end
return sum / (tp_conf.size * fp_conf.size).to_f
end
def f_measure(class_index=nil)
return prediction_feature_value_map( lambda{ |i| f_measure(i) } ) if class_index==nil
prec = precision(class_index)
rec = recall(class_index)
return 0 if prec == 0 and rec == 0
return 2 * prec * rec / (prec + rec).to_f;
end
def precision(class_index=nil)
return positive_predictive_value(class_index)
end
def positive_predictive_value(class_index=nil)
return prediction_feature_value_map( lambda{ |i| positive_predictive_value(i) } ) if class_index==nil
correct = 0 # all instances with prediction class_index that are correctly classified
total = 0 # all instances with prediciton class_index
(0..@num_classes-1).each do |i|
correct += @confusion_matrix[i][class_index] if i == class_index
total += @confusion_matrix[i][class_index]
end
return 0 if total==0
return correct/total.to_f
end
def negative_predictive_value(class_index=nil)
return prediction_feature_value_map( lambda{ |i| negative_predictive_value(i) } ) if class_index==nil
correct = 0 # all instances with prediction class_index that are correctly classified
total = 0 # all instances with prediciton class_index
(0..@num_classes-1).each do |i|
if i != class_index
(0..@num_classes-1).each do |j|
correct += @confusion_matrix[j][i] if j != class_index
total += @confusion_matrix[j][i]
end
end
end
return 0 if total==0
return correct/total.to_f
end
def recall(class_index=nil)
return true_positive_rate(class_index)
end
def true_negative_rate(class_index=nil)
return prediction_feature_value_map( lambda{ |i| true_negative_rate(i) } ) if class_index==nil
correct = 0
total = 0
(0..@num_classes-1).each do |i|
if i != class_index
(0..@num_classes-1).each do |j|
correct += @confusion_matrix[i][j] if j != class_index
total += @confusion_matrix[i][j]
end
end
end
return 0 if total==0
return correct/total.to_f
end
def num_true_negatives(class_index=nil)
return prediction_feature_value_map( lambda{ |i| num_true_negatives(i) } ) if class_index==nil
correct = 0
(0..@num_classes-1).each do |i|
if i != class_index
(0..@num_classes-1).each do |j|
correct += @confusion_matrix[i][j] if j != class_index
end
end
end
return correct
end
def true_positive_rate(class_index=nil)
return prediction_feature_value_map( lambda{ |i| true_positive_rate(i) } ) if class_index==nil
correct = 0
total = 0
(0..@num_classes-1).each do |i|
correct += @confusion_matrix[class_index][i] if i == class_index
total += @confusion_matrix[class_index][i]
end
return 0 if total==0
return correct/total.to_f
end
def num_true_positives(class_index=nil)
return prediction_feature_value_map( lambda{ |i| num_true_positives(i) } ) if class_index==nil
correct = 0
(0..@num_classes-1).each do |i|
correct += @confusion_matrix[class_index][i] if i == class_index
end
return correct
end
def false_negative_rate(class_index=nil)
return prediction_feature_value_map( lambda{ |i| false_negative_rate(i) } ) if class_index==nil
total = 0
incorrect = 0
(0..@num_classes-1).each do |i|
if i == class_index
(0..@num_classes-1).each do |j|
incorrect += @confusion_matrix[i][j] if j != class_index
total += @confusion_matrix[i][j]
end
end
end
return 0 if total == 0
return incorrect / total.to_f
end
def num_false_negatives(class_index=nil)
return prediction_feature_value_map( lambda{ |i| num_false_negatives(i) } ) if class_index==nil
incorrect = 0
(0..@num_classes-1).each do |i|
if i == class_index
(0..@num_classes-1).each do |j|
incorrect += @confusion_matrix[i][j] if j != class_index
end
end
end
return incorrect
end
def false_positive_rate(class_index=nil)
return prediction_feature_value_map( lambda{ |i| false_positive_rate(i) } ) if class_index==nil
total = 0
incorrect = 0
(0..@num_classes-1).each do |i|
if i != class_index
(0..@num_classes-1).each do |j|
incorrect += @confusion_matrix[i][j] if j == class_index
total += @confusion_matrix[i][j]
end
end
end
return 0 if total == 0
return incorrect / total.to_f
end
def num_false_positives(class_index=nil)
return prediction_feature_value_map( lambda{ |i| num_false_positives(i) } ) if class_index==nil
incorrect = 0
(0..@num_classes-1).each do |i|
if i != class_index
(0..@num_classes-1).each do |j|
incorrect += @confusion_matrix[i][j] if j == class_index
end
end
end
return incorrect
end
def average_area_under_roc
w_auc = average_measure( :area_under_roc )
w_auc.nan? ? 0 : w_auc
end
def average_f_measure
return average_measure( :f_measure )
end
private
# the <measure> is averaged over the number of instances for each actual class value
def average_measure( measure )
sum_instances = 0
num_instances_per_class = Array.new(@num_classes, 0)
(0..@num_classes-1).each do |i|
(0..@num_classes-1).each do |j|
num_instances_per_class[i] += @confusion_matrix[i][j]
end
sum_instances += num_instances_per_class[i]
end
raise "sum instances ("+sum_instances.to_s+") != num predicted ("+@num_predicted.to_s+")" unless @num_predicted == sum_instances
weighted = 0;
(0..@num_classes-1).each do |i|
weighted += self.send(measure,i) * num_instances_per_class[i]
end
return weighted / @num_predicted.to_f
end
# regression #######################################################################################
public
def root_mean_squared_error
return 0 if (@num_with_actual_value - @num_unpredicted)==0
mse = @sum_squared_error / (@num_with_actual_value - @num_unpredicted).to_f
return 0 if mse.nan?
Math.sqrt(mse)
end
def weighted_root_mean_squared_error
return 0 unless confidence_values_available?
return 0 if (@num_with_actual_value - @num_unpredicted)==0
Math.sqrt(@sum_weighted_squared_error / ((@num_with_actual_value - @num_unpredicted).to_f * @mean_confidence ))
end
def mean_absolute_error
return 0 if (@num_with_actual_value - @num_unpredicted)==0
@sum_abs_error / (@num_with_actual_value - @num_unpredicted).to_f
end
def weighted_mean_absolute_error
return 0 unless confidence_values_available?
return 0 if (@num_with_actual_value - @num_unpredicted)==0
@sum_weighted_abs_error / ((@num_with_actual_value - @num_unpredicted).to_f * @mean_confidence )
end
def sum_squared_error
return @sum_squared_error
end
def r_square #_old
#return sample_correlation_coefficient ** 2
# see http://en.wikipedia.org/wiki/Coefficient_of_determination#Definitions
# see http://web.maths.unsw.edu.au/~adelle/Garvan/Assays/GoodnessOfFit.html
ss_tot = total_sum_of_squares
return 0 if ss_tot==0
r_2 = 1 - residual_sum_of_squares / ss_tot
( r_2.infinite? || r_2.nan? ) ? 0 : r_2
end
def weighted_r_square #_old
return 0 unless confidence_values_available?
ss_tot = weighted_total_sum_of_squares
return 0 if ss_tot==0
r_2 = 1 - weighted_residual_sum_of_squares / ss_tot
( r_2.infinite? || r_2.nan? ) ? 0 : r_2
end
#def r_square
# # as implemted in R
# return sample_correlation_coefficient ** 2
#end
#def weighted_r_square
# # as implemted in R
# return weighted_sample_correlation_coefficient ** 2
#end
def sample_correlation_coefficient
begin
# formula see http://en.wikipedia.org/wiki/Correlation_and_dependence#Pearson.27s_product-moment_coefficient
scc = ( @num_predicted * @sum_multiply - @sum_actual * @sum_predicted ) /
( Math.sqrt( @num_predicted * @sum_squares_actual - @sum_actual**2 ) *
Math.sqrt( @num_predicted * @sum_squares_predicted - @sum_predicted**2 ) )
( scc.infinite? || scc.nan? ) ? 0 : scc
rescue; 0; end
end
def weighted_sample_correlation_coefficient
begin
# formula see http://en.wikipedia.org/wiki/Correlation_and_dependence#Pearson.27s_product-moment_coefficient
scc = ( @num_predicted * @weighted_sum_multiply - @weighted_sum_actual * @weighted_sum_predicted ) /
( Math.sqrt( @num_predicted * @weighted_sum_squares_actual - @weighted_sum_actual**2 ) *
Math.sqrt( @num_predicted * @weighted_sum_squares_predicted - @weighted_sum_predicted**2 ) )
( scc.infinite? || scc.nan? ) ? 0 : scc
rescue; 0; end
end
def total_sum_of_squares
#return @variance_actual * ( @num_predicted - 1 )
sum = 0
@predicted_values.size.times do |i|
sum += (@actual_values[i]-@actual_mean)**2 if @actual_values[i]!=nil and @predicted_values[i]!=nil
end
sum
end
def weighted_total_sum_of_squares
return 0 unless confidence_values_available?
sum = 0
@predicted_values.size.times do |i|
sum += ((@actual_values[i]-@actual_mean)**2)*@confidence_values[i] if @actual_values[i]!=nil and @predicted_values[i]!=nil
end
sum
end
def residual_sum_of_squares
sum_squared_error
end
def weighted_residual_sum_of_squares
@sum_weighted_squared_error
end
def target_variance_predicted
return @variance_predicted
end
def target_variance_actual
return @variance_actual
end
# data for (roc-)plots ###################################################################################
def get_roc_prediction_values(class_value)
#puts "get_roc_values for class_value: "+class_value.to_s
raise "no confidence values" unless confidence_values_available?
raise "no class-value specified" if class_value==nil
class_index = @accept_values.index(class_value) if class_value!=nil
raise "class not found "+class_value.to_s if (class_value!=nil && class_index==nil)
c = []; tp = []
(0..@predicted_values.size-1).each do |i|
if @predicted_values[i]!=nil
c << @confidence_values[i] * (@predicted_values[i]==class_index ? 1 : -1)
if (@actual_values[i]==class_index)
tp << 1
else
tp << 0
end
end
end
# DO NOT raise exception here, maybe different validations are concated
#raise "no instance predicted as '"+class_value+"'" if p.size == 0
h = {:true_positives => tp, :confidence_values => c}
#puts h.inspect
return h
end
def get_prediction_values(performance_attr, performance_accept_value)
#puts "get_roc_values for class_value: "+class_value.to_s
raise "no confidence values" unless confidence_values_available?
#raise "no class-value specified" if class_value==nil
actual_accept_value = nil
predicted_accept_value = nil
if performance_attr==:true_positive_rate
actual_accept_value = performance_accept_value
elsif performance_attr==:positive_predictive_value
predicted_accept_value = performance_accept_value
end
actual_class_index = @accept_values.index(actual_accept_value) if actual_accept_value!=nil
raise "class not found '"+actual_accept_value.to_s+"' in "+@accept_values.inspect if (actual_accept_value!=nil && actual_class_index==nil)
predicted_class_index = @accept_values.index(predicted_accept_value) if predicted_accept_value!=nil
raise "class not found '"+predicted_accept_value.to_s+"' in "+@accept_values.inspect if (predicted_accept_value!=nil && predicted_class_index==nil)
c = []; p = []; a = []
(0..@predicted_values.size-1).each do |i|
# NOTE: not predicted instances are ignored here
if @predicted_values[i]!=nil and
(predicted_class_index==nil || @predicted_values[i]==predicted_class_index) and
(actual_class_index==nil || @actual_values[i]==actual_class_index)
c << @confidence_values[i]
p << @predicted_values[i]
a << @actual_values[i]
end
end
# DO NOT raise exception here, maybe different validations are concated
#raise "no instance predicted as '"+class_value+"'" if p.size == 0
h = {:predicted_values => p, :actual_values => a, :confidence_values => c}
#puts h.inspect
return h
end
########################################################################################
def num_instances
return @predicted_values.size
end
def predicted_values
@predicted_values
end
def predicted_value(instance_index)
case @feature_type
when "classification"
@predicted_values[instance_index]==nil ? nil : @accept_values[@predicted_values[instance_index]]
when "regression"
@predicted_values[instance_index]
end
end
def actual_values
@actual_values
end
def actual_value(instance_index)
case @feature_type
when "classification"
@actual_values[instance_index]==nil ? nil : @accept_values[@actual_values[instance_index]]
when "regression"
@actual_values[instance_index]
end
end
def confidence_value(instance_index)
return @confidence_values[instance_index]
end
def classification_miss?(instance_index)
raise "no classification" unless @feature_type=="classification"
return false if predicted_value(instance_index)==nil or actual_value(instance_index)==nil
return predicted_value(instance_index) != actual_value(instance_index)
end
def feature_type
@feature_type
end
def confidence_values_available?
@conf_provided
end
def min_confidence
@confidence_values[-1]
end
###################################################################################################################
#def compound(instance_index)
#return "instance_index.to_s"
#end
private
def self.test_update
p=[0.4,0.2,0.3,0.5,0.8]
a=[0.45,0.21,0.25,0.55,0.75]
c = Array.new(p.size)
pred = Predictions.new(p,a,c,"regression")
puts pred.r_square
pred = nil
p.size.times do |i|
if pred==nil
pred = Predictions.new([p[0]],[a[0]],[c[0]],"regression")
else
pred.update_stats(p[i],a[i],c[i])
end
puts pred.r_square
end
end
def self.test_r_square
require "rubygems"
require "opentox-ruby"
max_deviation = rand * 0.9
avg_deviation = max_deviation * 0.5
p = []
a = []
c = []
(100 + rand(1000)).times do |i|
r = rand
deviation = rand * max_deviation
a << r
p << r + ((rand<0.5 ? -1 : 1) * deviation)
#c << 0.5
if (deviation > avg_deviation)
c << 0.4
else
c << 0.6
end
#puts a[-1].to_s+" "+p[-1].to_s
end
puts "num values "+p.size.to_s
pred = Predictions.new(p,a,c,"regression")
puts "internal"
#puts "r-square old "+pred.r_square_old.to_s
puts "cor "+pred.sample_correlation_coefficient.to_s
puts "weighted cor "+pred.weighted_sample_correlation_coefficient.to_s
puts "r-square "+pred.r_square.to_s
puts "R"
@@r = RinRuby.new(true,false) unless defined?(@@r) and @@r
@@r.assign "v1",a
@@r.assign "v2",p
puts "r cor "+@@r.pull("cor(v1,v2)").to_s
@@r.eval "fit <- lm(v1 ~ v2)"
@@r.eval "sum <- summary(fit)"
puts "r r-square "+@@r.pull("sum$r.squared").to_s
puts "r adjusted-r-square "+@@r.pull("sum$adj.r.squared").to_s
end
def prediction_feature_value_map(proc)
res = {}
(0..@num_classes-1).each do |i|
res[@accept_values[i]] = proc.call(i)
end
return res
end
end
end
#class Float
# def to_s
# "%.5f" % self
# end
#end
##Lib::Predictions.test_update
#Lib::Predictions.test_r_square
|