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
|
module OpenTox
module Validation
# Statistical evaluation of classification validations
module ClassificationStatistics
# Get statistics
# @return [Hash]
def statistics
self.accept_values = model.prediction_feature.accept_values
self.confusion_matrix = {:all => Array.new(accept_values.size){Array.new(accept_values.size,0)}, :without_warnings => Array.new(accept_values.size){Array.new(accept_values.size,0)}}
self.weighted_confusion_matrix = {:all => Array.new(accept_values.size){Array.new(accept_values.size,0)}, :without_warnings => Array.new(accept_values.size){Array.new(accept_values.size,0)}}
#self.weighted_confusion_matrix = Array.new(accept_values.size){Array.new(accept_values.size,0)}
self.nr_predictions = {:all => 0,:without_warnings => 0}
predictions.each do |cid,pred|
# TODO
# use predictions without probabilities (single neighbor)??
# use measured majority class??
if pred[:measurements].uniq.size == 1 and pred[:probabilities]
m = pred[:measurements].first
if pred[:value] == m
if pred[:value] == accept_values[0]
confusion_matrix[:all][0][0] += 1
weighted_confusion_matrix[:all][0][0] += pred[:probabilities][pred[:value]]
self.nr_predictions[:all] += 1
if pred[:warnings].empty?
confusion_matrix[:without_warnings][0][0] += 1
weighted_confusion_matrix[:without_warnings][0][0] += pred[:probabilities][pred[:value]]
self.nr_predictions[:without_warnings] += 1
end
elsif pred[:value] == accept_values[1]
confusion_matrix[:all][1][1] += 1
weighted_confusion_matrix[:all][1][1] += pred[:probabilities][pred[:value]]
self.nr_predictions[:all] += 1
if pred[:warnings].empty?
confusion_matrix[:without_warnings][1][1] += 1
weighted_confusion_matrix[:without_warnings][1][1] += pred[:probabilities][pred[:value]]
self.nr_predictions[:without_warnings] += 1
end
end
elsif pred[:value] != m
if pred[:value] == accept_values[0]
confusion_matrix[:all][0][1] += 1
weighted_confusion_matrix[:all][0][1] += pred[:probabilities][pred[:value]]
self.nr_predictions[:all] += 1
if pred[:warnings].empty?
confusion_matrix[:without_warnings][0][1] += 1
weighted_confusion_matrix[:without_warnings][0][1] += pred[:probabilities][pred[:value]]
self.nr_predictions[:without_warnings] += 1
end
elsif pred[:value] == accept_values[1]
confusion_matrix[:all][1][0] += 1
weighted_confusion_matrix[:all][1][0] += pred[:probabilities][pred[:value]]
self.nr_predictions[:all] += 1
if pred[:warnings].empty?
confusion_matrix[:without_warnings][1][0] += 1
weighted_confusion_matrix[:without_warnings][1][0] += pred[:probabilities][pred[:value]]
self.nr_predictions[:without_warnings] += 1
end
end
end
end
end
self.true_rate = {:all => {}, :without_warnings => {}}
self.predictivity = {:all => {}, :without_warnings => {}}
accept_values.each_with_index do |v,i|
[:all,:without_warnings].each do |a|
self.true_rate[a][v] = confusion_matrix[a][i][i]/confusion_matrix[a][i].reduce(:+).to_f
self.predictivity[a][v] = confusion_matrix[a][i][i]/confusion_matrix[a].collect{|n| n[i]}.reduce(:+).to_f
end
end
confidence_sum = {:all => 0, :without_warnings => 0}
[:all,:without_warnings].each do |a|
weighted_confusion_matrix[a].each do |r|
r.each do |c|
confidence_sum[a] += c
end
end
end
self.accuracy = {}
self.weighted_accuracy = {}
[:all,:without_warnings].each do |a|
self.accuracy[a] = (confusion_matrix[a][0][0]+confusion_matrix[a][1][1])/nr_predictions[a].to_f
self.weighted_accuracy[a] = (weighted_confusion_matrix[a][0][0]+weighted_confusion_matrix[a][1][1])/confidence_sum[a].to_f
end
$logger.debug "Accuracy #{accuracy}"
save
{
:accept_values => accept_values,
:confusion_matrix => confusion_matrix,
:weighted_confusion_matrix => weighted_confusion_matrix,
:accuracy => accuracy,
:weighted_accuracy => weighted_accuracy,
:true_rate => self.true_rate,
:predictivity => self.predictivity,
:nr_predictions => nr_predictions,
}
end
# Plot accuracy vs prediction probability
# @param [String,nil] format
# @return [Blob]
def probability_plot format: "pdf"
#unless probability_plot_id
#tmpdir = File.join(ENV["HOME"], "tmp")
tmpdir = "/tmp"
#p tmpdir
FileUtils.mkdir_p tmpdir
tmpfile = File.join(tmpdir,"#{id.to_s}_probability.#{format}")
accuracies = []
probabilities = []
correct_predictions = 0
incorrect_predictions = 0
pp = []
predictions.values.select{|p| p["probabilities"]}.compact.each do |p|
p["measurements"].each do |m|
pp << [ p["probabilities"][p["value"]], p["value"] == m ]
end
end
pp.sort_by!{|p| 1-p.first}
pp.each do |p|
p[1] ? correct_predictions += 1 : incorrect_predictions += 1
accuracies << correct_predictions/(correct_predictions+incorrect_predictions).to_f
probabilities << p[0]
end
R.assign "accuracy", accuracies
R.assign "probability", probabilities
R.eval "image = qplot(probability,accuracy)+ylab('Accumulated accuracy')+xlab('Prediction probability')+ylim(c(0,1))+scale_x_reverse()+geom_line()"
R.eval "ggsave(file='#{tmpfile}', plot=image)"
file = Mongo::Grid::File.new(File.read(tmpfile), :filename => "#{self.id.to_s}_probability_plot.svg")
plot_id = $gridfs.insert_one(file)
update(:probability_plot_id => plot_id)
#end
$gridfs.find_one(_id: probability_plot_id).data
end
end
# Statistical evaluation of regression validations
module RegressionStatistics
# Get statistics
# @return [Hash]
def statistics
self.warnings = []
self.rmse = {:all =>0,:without_warnings => 0}
self.mae = {:all =>0,:without_warnings => 0}
self.within_prediction_interval = {:all =>0,:without_warnings => 0}
self.out_of_prediction_interval = {:all =>0,:without_warnings => 0}
x = {:all => [],:without_warnings => []}
y = {:all => [],:without_warnings => []}
self.nr_predictions = {:all =>0,:without_warnings => 0}
error = {}
predictions.each do |cid,pred|
if pred[:value] and pred[:measurements]
self.nr_predictions[:all] +=1
x[:all] << pred[:measurements].median
y[:all] << pred[:value]
error[:all] = pred[:value]-pred[:measurements].median
self.rmse[:all] += error**2
self.mae[:all] += error.abs
if pred[:prediction_interval]
if pred[:measurements].median >= pred[:prediction_interval][0] and pred[:measurements].median <= pred[:prediction_interval][1]
self.within_prediction_interval[:all] += 1
else
self.out_of_prediction_interval[:all] += 1
end
end
if pred[:warnings].empty?
self.nr_predictions[:without_warnings] +=1
x[:without_warnings] << pred[:measurements].median
y[:without_warnings] << pred[:value]
error[:without_warnings] = pred[:value]-pred[:measurements].median
self.rmse[:without_warnings] += error**2
self.mae[:without_warnings] += error.abs
if pred[:prediction_interval]
if pred[:measurements].median >= pred[:prediction_interval][0] and pred[:measurements].median <= pred[:prediction_interval][1]
self.within_prediction_interval[:without_warnings] += 1
else
self.out_of_prediction_interval[:without_warnings] += 1
end
end
end
else
trd_id = model.training_dataset_id
smiles = Compound.find(cid).smiles
self.warnings << "No training activities for #{smiles} in training dataset #{trd_id}."
$logger.debug "No training activities for #{smiles} in training dataset #{trd_id}."
end
end
[:all,:without_warnings].each do |a|
R.assign "measurement", x[a]
R.assign "prediction", y[a]
R.eval "r <- cor(measurement,prediction,use='pairwise')"
self.r_squared[a] = R.eval("r").to_ruby**2
self.mae[a] = self.mae[a]/self.nr_predictions[a]
self.rmse[a] = Math.sqrt(self.rmse[a]/self.nr_predictions[a])
end
$logger.debug "R^2 #{r_squared}"
$logger.debug "RMSE #{rmse}"
$logger.debug "MAE #{mae}"
$logger.debug "#{percent_within_prediction_interval.round(2)}% of measurements within prediction interval"
$logger.debug "#{warnings}"
save
{
:mae => mae,
:rmse => rmse,
:r_squared => r_squared,
:within_prediction_interval => within_prediction_interval,
:out_of_prediction_interval => out_of_prediction_interval,
:nr_predictions => nr_predictions,
}
end
# Get percentage of measurements within the prediction interval
# @return [Float]
def percent_within_prediction_interval
100*within_prediction_interval.to_f/(within_prediction_interval+out_of_prediction_interval)
end
# Plot predicted vs measured values
# @param [String,nil] format
# @return [Blob]
def correlation_plot format: "png"
unless correlation_plot_id
tmpfile = "/tmp/#{id.to_s}_correlation.#{format}"
x = []
y = []
feature = Feature.find(predictions.first.last["prediction_feature_id"])
predictions.each do |sid,p|
x << p["measurements"].median
y << p["value"]
end
R.assign "measurement", x
R.assign "prediction", y
R.eval "all = c(measurement,prediction)"
R.eval "range = c(min(all), max(all))"
if feature.name.match /Net cell association/ # ad hoc fix for awkward units
title = "log2(Net cell association [mL/ug(Mg)])"
else
title = feature.name
title += " [#{feature.unit}]" if feature.unit and !feature.unit.blank?
end
R.eval "image = qplot(prediction,measurement,main='#{title}',xlab='Prediction',ylab='Measurement',asp=1,xlim=range, ylim=range)"
R.eval "image = image + geom_abline(intercept=0, slope=1)"
R.eval "ggsave(file='#{tmpfile}', plot=image)"
file = Mongo::Grid::File.new(File.read(tmpfile), :filename => "#{id.to_s}_correlation_plot.#{format}")
plot_id = $gridfs.insert_one(file)
update(:correlation_plot_id => plot_id)
end
$gridfs.find_one(_id: correlation_plot_id).data
end
# Get predictions with measurements outside of the prediction interval
# @return [Hash]
def worst_predictions
worst_predictions = predictions.select do |sid,p|
p["prediction_interval"] and p["value"] and (p["measurements"].max < p["prediction_interval"][0] or p["measurements"].min > p["prediction_interval"][1])
end.compact.to_h
worst_predictions.each do |sid,p|
p["error"] = (p["value"] - p["measurements"].median).abs
if p["measurements"].max < p["prediction_interval"][0]
p["distance_prediction_interval"] = (p["measurements"].max - p["prediction_interval"][0]).abs
elsif p["measurements"].min > p["prediction_interval"][1]
p["distance_prediction_interval"] = (p["measurements"].min - p["prediction_interval"][1]).abs
end
end
worst_predictions.sort_by{|sid,p| p["distance_prediction_interval"] }.to_h
end
end
end
end
|
