File size: 16,754 Bytes
eb90a8a | 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 | """
Module for constructing and analysing confusion matrix
of source code classification
Prints confusion report and writes down
statistics of classification accuracy
"""
import sys
import os
from abc import ABC, abstractmethod
import numpy as np
import csv
import sklearn.metrics as metrics
from ConfusionAnalysis import ConfusionAnalysis
class ClassConfusAnalysis(ConfusionAnalysis):
"""
Class for constructing and analysing confusion matrix
"""
def __init__(self, probabilities, labels, solutions,
problems, extreme = 0.1):
"""
Construct confusion matrix
Parameters:
- probabilities - classification probabilities as numpy array
- labels - list of samples labels
- solutions - list of file names of problem solutions,
i.e sample names for this case
- problems - list of problems
i.e names of classes for this case
- extreme - fraction of cases to be reported
as extreme ones
"""
super(ClassConfusAnalysis, self).__init__(probabilities,
np.asarray(labels, dtype = np.int32),
solutions, problems, extreme)
self.class_stat_fn = "class_statistics.lst"
self.class_stat_csv = "class_statistics.csv"
self.confused_classes_fn = "confused_classes.lst"
self.confused_classes_csv = "confused_classes.csv"
#Number of best and worst correct predictions of
#worst classified classes to report
self.n_right_pred_report = 32
#Max number of confusions to report
self.max_n_confus_report = 64
#Max number of worst and best classes to report mistakes
self.max_n_class_report = 16
#Max number of misclassifications to report
self.max_n_misclass_report = 32
#Confidence in classification
self.confidence = np.max(probabilities, axis=1)
#Number of samples in each class as numpy array
self._n_class_samples = self.conf_mat.sum(axis=1)
#Number of predictions of each class as numpy array
self._n_class_predict = self.conf_mat.sum(axis=0)
#Number of correct predictions of each class
self.n_corr_class_samples = np.diag(self.conf_mat)
#Confusion matrix normalized with number of class samples
self.norm_conf_mat = \
self.conf_mat / (self._n_class_samples[:, np.newaxis])
#Accuracy of class predictions as python list
self.class_accuracy = \
np.diag(self.norm_conf_mat).tolist()
def getPredictions(self):
"""
Get classification predictions
Returns: classification predictions
"""
return np.argmax(self.probabilities, axis=1)
def reportClasses(self, classes, f):
"""
Report classification of set of classes
Parameters:
- classes -- classes to report
- f -- file to write report to
"""
self.reportClassProblems(self._n_class_samples,
self.n_corr_class_samples,
classes, f)
self.csvClassProblems(self._n_class_samples,
self.n_corr_class_samples,
classes)
def csvClassProblems(self, tot_samples, corr_samples,
problems):
"""
Writes absolute and relative numbers of
correct and incorrect classifications of problems
to csv file
Parameters:
- tot_samples -- vector of total numbers of samples
corresponding to each problem
- corr_samples -- vector of numbers of correctly classified samples
- problems -- list of indices of problems to report
"""
with open(f"{self.report_dir}/{self.class_stat_csv}",
'w', newline='') as _csv_f:
_w = csv.writer(_csv_f)
_w.writerow(("Label", "Accuracy", "N files",
"N errors", "Problem"))
for _c in problems:
_w.writerow((_c,
100.0 * float(corr_samples[_c]) / float(tot_samples[_c]),
tot_samples[_c], tot_samples[_c] - corr_samples[_c],
self.problems[_c]))
def reportClassProblems(self, tot_samples, corr_samples,
problems, f):
"""
Writes absolute and relative numbers of
correct and incorrect classifications of problems
Parameters:
- tot_samples -- vector of total numbers of samples
corresponding to each problem
- corr_samples -- vector of numbers of correctly classified samples
- problems -- list of indices of problems to report
- f -- file to write to
"""
f.write("Label Accur Size N errs Problem name\n")
for _c in sorted(problems,
key = lambda _p: float(corr_samples[_p]) / float(tot_samples[_p])):
f.write("{:4d} {:5.1f}% {:5d} {:5d} {:s}\n".
format(_c,
100.0 * float(corr_samples[_c]) / float(tot_samples[_c]),
tot_samples[_c], tot_samples[_c] - corr_samples[_c],
self.problems[_c]))
def getRightClassifications(self, classes):
"""
Get list of correct classifications of given classes
Parameters:
- classes -- indices of classes to report
Returns list of lists of pairs, each of which is
<index of classified sample, confidence>
Sublists correspond to given classes
Each sublist is sorted according to confidence
"""
_classifications = [[] for i in range(len(classes))]
for _i in range(self._n_samples):
if (self.predictions[_i] == self.labels[_i] and
self.labels[_i] in classes):
_idx = classes.index(self.labels[_i])
_classifications[_idx].append((self.solutions[_i],
self.confidence[_i]))
for _cls in _classifications:
if len(_cls) < 2: continue
_cls.sort(key = lambda _c: _c[1], reverse = True)
return _classifications
def reportRightClassifications(self, classes, f):
"""
Report correct classifications of given classes
Parameters:
- classes -- indices of classes to report
- f -- file to write to
"""
_classifications = self.getRightClassifications(classes)
for _i, _c in enumerate(sorted(classes,
key = lambda _cl: self.class_accuracy[_cl])):
if not _classifications[_i]: continue
f.write(f"#{_i} Problem {self.problems[_c]} number {_c} " +
f"is classified with {100.0 *self.class_accuracy[_c]:.2f}% " +
"accuracy\n")
f.write(f"There are {len(_classifications[_i])} correct classifications\n")
if len(_classifications[_i]) < 2 * self.n_right_pred_report + 3:
self.writeClassifications(_classifications[_i],
"Solution predictions", f)
else:
self.writeClassifications(
_classifications[_i][: self.n_right_pred_report],
"Strongest predictions", f)
self.writeClassifications(
_classifications[_i][-self.n_right_pred_report :],
"Weakest predictions", f)
def writeClassifications(self, predictions, title, f):
"""
Write down information on classification predictions
Parameters:
- predictions - list of predictions each of which is
a pair <index of classified sample, confidence>
- title - text to put in the first line
- f - file to write to
"""
_line = title
_punct = ': '
for _pred in predictions:
_line += _punct
_punct = ', '
if len(_line) > 80:
f.write(_line + "\n")
_line = ""
_line += f"{_pred[0]}/{100 * _pred[1]:.2f}%"
if _line: f.write(_line + "\n")
def getClassMistakes(self, classes):
"""
Get list of misclassifications for given classes
Parameters:
- classes -- indices of classes to report
Returns list of lists of miscalssifications:
pairs <index of classified sample, prediction>
Sublists gives misclassifications of classes
"""
#Number of classification mistakes
_mistakes = [[] for i in range(len(classes))]
for _i in range(self._n_samples):
if (self.predictions[_i] != self.labels[_i] and
self.labels[_i] in classes):
_bad_idx = classes.index(self.labels[_i])
_mistakes[_bad_idx].append((_i, self.predictions[_i]))
return _mistakes
def reportMisclassifications(self, classes, mistakes, f):
"""
Report misclassifications of given classes
Parameters:
- classes -- indices of classes to report
- mistakes -- list of lists of miscalssifications:
pairs <classified class index, prediction>
Sublists gives misclassifications of classes
- f -- file to write to
"""
for _i, _c in enumerate(sorted(classes,
key = lambda _cl: self.class_accuracy[_cl])):
if not mistakes[_i]: continue
f.write(f"#{_i} Problem {self.problems[_c]} number {_c} " +
f"is classified with {100.0 *self.class_accuracy[_c]:.2f}% " +
"accuracy\n")
_line = f"There are {len(mistakes[_i])} misclassified solutions"
_punct = ': '
for _err in sorted(mistakes[_i][: self.max_n_misclass_report],
key = lambda _e: self.confidence[_e[0]],
reverse = True):
_conf = "{:5.2f}".format(100.0 * self.confidence[_err[0]])
_line += _punct
_punct = ', '
if len(_line) > 80:
f.write(_line + "\n")
_line = ""
_line += (self.solutions[_err[0]] + " => " +
self.problems[_err[1]] + ' (' + _conf + '%)')
if _line: f.write(_line + "\n")
def reportConfusedClasses(self):
"""
Report worst classification confusions,
i.e. confused pairs of confused problems
"""
_confusions = self.worstConfusions()
_confusions = [_c for _c in _confusions if _c[0] != _c[1]]
with open(f"{self.report_dir}/{self.confused_classes_fn}", 'w') as _lst, \
open(f"{self.report_dir}/{self.confused_classes_csv}",
'w', newline='') as _csv:
_lst.write("Most confused pairs of problems\n")
_lst.write("-------------------------------\n")
_lst.write(
" Orig Pred N N %% Original Predicted\n")
_lst.write(
"Label Label Samples Errors Errors Problem Problem \n")
_w = csv.writer(_csv)
_w.writerow(("Original_Label", "Predicted_Label",
"N_samples",
"%%_of_errors", "Original_Problem",
"Predicted Problem"))
for _c in sorted(_confusions,
key = lambda _conf:
self.norm_conf_mat[_conf[0], _conf[1]],
reverse = True):
_lorig, _lpred = _c
_lst.write("{:5d} {:5d} {:6d} {:6d} {:6.2f} {:12s} {:12s}\n".
format(_lorig, _lpred,
self._n_class_samples[_lorig],
self.conf_mat[_lorig, _lpred],
self.norm_conf_mat[_lorig, _lpred] * 100,
self.problems[_lorig], self.problems[_lpred]))
_w.writerow((_lorig, _lpred, self._n_class_samples[_lorig],
self.norm_conf_mat[_lorig, _lpred] * 100,
self.problems[_lorig], self.problems[_lpred]))
def worstConfusions(self):
"""
Compute worst confusions of classification
Returns list of pairs , each of which is
a classified problem and the problem predicted
"""
_n_confusions_report = \
max(min(int(self._n_problems * self._n_problems *
self._n_extreme_cases), self.max_n_confus_report),
self._n_problems)
_flat_conf = self.norm_conf_mat.flatten()
_part_index = \
self._n_problems * self._n_problems
_part_index -= (_n_confusions_report + self._n_problems)
_worst_idx = \
np.argpartition(_flat_conf, _part_index)[_part_index :]
_worst_idx = \
np.unravel_index(_worst_idx,
self.norm_conf_mat.shape)
_worst_idx = zip(_worst_idx[0].tolist(), _worst_idx[1].tolist())
return _worst_idx
def printClassAccuracy(self, f):
"""
Report accuracy of classifications
and their misclassified samples
Parameters:
-f -- file to write report to
"""
f.write("\nClassification statistics\n")
f.write("-------------------------\n")
self.reportClasses(range(self._n_problems), f)
_class_acc = \
sorted(zip(range(self._n_problems), self.class_accuracy),
key = lambda _s : _s[1])
_n_class_report = min(self._n_problems // 2,
self.max_n_class_report)
_classes = _class_acc[: _n_class_report]
_worst_classes = list(zip(*_classes))[0]
_classes = _class_acc[-_n_class_report :]
_best_classes = list(zip(*_classes))[0]
_mistakes = self.getClassMistakes(_worst_classes + _best_classes)
f.write("\nWorst classified classes\n")
f.write("--------------------------\n")
self.reportClasses(_worst_classes, f)
f.write("\nMisclassifications of the worst classified classes:\n")
f.write("-----------------------------------------------------\n")
self.reportMisclassifications(_worst_classes,
_mistakes[: len(_worst_classes)], f)
f.write("\nCorrect classifications of the worst classified classes:\n")
f.write("-----------------------------------------------------\n")
self.reportRightClassifications(_worst_classes, f)
f.write("\nBest classified classes\n")
f.write("-------------------------\n")
self.reportClasses(_best_classes, f)
f.write("\nMisclassifications of the best classified classes:\n")
f.write("---------------------------------------------------\n")
self.reportMisclassifications(_best_classes,
_mistakes[len(_worst_classes) :], f)
def writeReport(self):
"""
Write report on classification and misclassification
"""
with open(f"{self.report_dir}/{self._conf_mat_fn}", 'w') as _f:
_f.write("Unnormalized confusion matrix\n")
_f.write("row and column indices are labels\n")
_f.write("---------------------------------\n")
self.writeLargeMatrixInt(self.conf_mat, _f)
_f.write("Normalized confusion matrix\n")
_f.write("row and column indices are labels\n")
_f.write("---------------------------------\n")
self.writeLargeMatrixPct(self.norm_conf_mat, _f, 100.0)
with open(f"{self.report_dir}/{self.class_stat_fn}", 'w') as _f:
self.printClassAccuracy(_f)
self.reportConfusedClasses()
print("\nClassification report")
print("-----------------------")
print(metrics.classification_report(self.labels,
self.predictions,
target_names = self.problems))
#---------------- End of class ClassConfusAnalysis ----------------------
|