File size: 23,615 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 | """
Module for analysing confusion matrix and accuracy
of similarity analysis of source code
"""
import sys
import os
from abc import ABC, abstractmethod
import pickle
import numpy as np
import sklearn.metrics as metrics
from ConfusionAnalysis import ConfusionAnalysis
class SimilConfusAnalysis(ConfusionAnalysis):
"""
Class for constructing and analysing confusion matrix
"""
@classmethod
def similarityToProb(cls, sim_metric, labels01):
"""
Convert similarity metric to probabilities
Parameters:
- sim_metric -- similarity metric
- labels01 -- label types flag:
True: 0/1 labels
Flase: -1/+1 labels
"""
if labels01:
return sim_metric
return (sim_metric + 1.0) / 2.0
def __init__(self, probabilities, labels, solutions,
problems, annotations, extreme = 0.1,
labels01 = True):
"""
Construct confusion matrix
Parameters:
- probabilities - classification probabilities as numpy array
- labels - labels of samples as numpy array
- solutions - list of sub-lists of file names of problem solutions
Each sub-list contains sample names of one problem
All sublists are part of the whole list of sample names
- problems - list of names of problems
- annotations - list of sample annotations
Each sample is represented as 4-tuple
<problem 1, solution 1, problem 2, solution 2>,
where problems and solutions are their indices
- extreme - fraction of cases to be reported
as extreme ones
- labels01 -- label types flag:
True: 0/1 labels
Flase: -1/+1 labels
"""
super(SimilConfusAnalysis, self).__init__(
self.similarityToProb(probabilities, labels01),
self.similarityToProb(labels, labels01),
solutions, problems, extreme)
self.annotations = annotations
self.simil_stat_fn = "similarity_stat.lst"
self.dissimil_stat_fn = "dissimilarity_stat.lst"
self.max_n_cases_report = 32
self._n_cases_report = min(self.max_n_cases_report, self._n_extreme_cases)
self.tn, self.fp, self.fn, self.tp = self.conf_mat.ravel()
self.precision_score = \
metrics.precision_score(self.labels, self.predictions)
self.recall_score = \
metrics.recall_score(self.labels, self.predictions)
self.f1_score = \
metrics.f1_score(self.labels, self.predictions)
self.fbeta_score_05 = \
metrics.fbeta_score(self.labels, self.predictions, beta=0.5)
self.fbeta_score_1 = \
metrics.fbeta_score(self.labels, self.predictions, beta=1)
self.fbeta_score_2 = \
metrics.fbeta_score(self.labels, self.predictions, beta=2)
self.precision, self.recall, self.threshold = \
metrics.precision_recall_curve(self.labels, self.probabilities)
self.average_precision_score = \
metrics.average_precision_score(self.labels, self.probabilities)
def getPredictions(self):
"""
Get classification predictions
Returns: - classification predictions
"""
return self.probabilities.round()[:,0].astype(int)
def printConfusionReport(self):
""""
Print confusion report
"""
print("\nClassification report")
print("-----------------------")
print(metrics.classification_report(
self.labels, self.predictions,
target_names = ["dissimilar", "similar"]))
print("\nConfusion matrix")
print("----------------")
print(" Dissimilar Similar")
print("Dissimilar {:7d} {:7d}".
format(self.conf_mat[0, 0], self.conf_mat[0,1]))
print("Similar {:7d} {:7d}".
format(self.conf_mat[1, 0], self.conf_mat[1, 1]))
print("\n")
print(f"tn = {self.tn}, fp = {self.fp}," +
f"fn = {self.fn}, tp = {self.tp}")
print(f"precision_score = {self.precision_score:.4f}")
print(f"recall_score = {self.recall_score:.4f}")
print(f"f1_score = {self.f1_score:.4f}")
print(f"fbeta_score(0.5) = {self.fbeta_score_05:.4f}")
print(f"fbeta_score(1) = {self.fbeta_score_1:.4f}")
print(f"fbeta_score(2) = {self.fbeta_score_2:.4f}")
print(f"average_precision_score = {self.average_precision_score:.4f}")
with open("roc.pcl", 'wb') as _jar:
_roc = {"precision" : self.precision,
"recall" : self.recall,
"threshold" : self.threshold}
pickle.dump(_roc, _jar)
def similarityConfusion(self):
"""
Compute and print following matricies:
- matrix of numbers of similarity/dissimilarity tests
for solutions of pairs of problems
- matrix of correct detections of similarity/dissimilarity
of solutions of pairs of problems
"""
#Matrix of numbers of cases when similarity or dissimilarity
#of two solutions is detected correctly
self.sim_num_correct = \
np.zeros((self._n_problems, self._n_problems), dtype=int)
#Matrix of total numbers of similarity/dissimilarity tests
self.sim_num_samples = \
np.zeros((self._n_problems, self._n_problems), dtype=int)
for _s, _p, _l in zip(self.annotations, self.predictions,
self.labels):
_problem1, _, _problem2, _ = _s
self.sim_num_samples[_problem1, _problem2] += 1
self.sim_num_correct[_problem1, _problem2] += int(_p == _l)
def compSimTestAccuracy(self):
"""
Compute lists of accuracy of similarity and dissimilarity detection
Returns:
- list of tested problems
i-th element of list is either True or False
indicating if solutions of i-th problem were tested
- list of accuracy of similarity detection
An element of the list is a pair: <problem, accuracy>
- list of accuracy of dissimilarity tests
Elements of this list are tuples
<problem1, problem2, acc1, acc2, acc1+acc2, abs(acc1-acc2)>
If any accuracy acc1(acc2) is not defined:
acc1(acc2) = None
acc1+acc2 = 2*acc2(2*acc1)
abs(acc1-acc2) = 0
"""
_sim = []
_dissim = []
_tested_problems = [False] * len(self.problems)
for _i in range(self._n_problems):
for _j in range(self._n_problems):
if _j > _i: break
_ntests_ij = self.sim_num_samples[_i, _j]
_ntests_ji = self.sim_num_samples[_j, _i]
if _ntests_ij + _ntests_ji == 0: continue
_tested_problems[_i] = True
_tested_problems[_j] = True
if _j == _i:
_acc = self.sim_num_correct[_i, _j] / _ntests_ij
_sim.append((_i, _acc))
continue
if _ntests_ij == 0:
_acc_ji = self.sim_num_correct[_j, _i] / _ntests_ji
_dissim.append((_j, _i, None, _acc_ji, _acc_ji, 0))
continue
if _ntests_ji == 0:
_acc_ij = self.sim_num_correct[_i, _j] / _ntests_ij
_dissim.append((_i, _j, _acc_ij, None, _acc_ij, 0))
continue
_acc_ij = self.sim_num_correct[_i, _j] / _ntests_ij
_acc_ji = self.sim_num_correct[_j, _i] / _ntests_ji
_acc = ((self.sim_num_correct[_i, _j] +
self.sim_num_correct[_j, _i]) /
(_ntests_ij + _ntests_ji))
_dissim.append((_i, _j, _acc_ij, _acc_ji, _acc,
abs(_acc_ij - _acc_ji)))
return _tested_problems, _sim, _dissim
def writeSimDissimAccuracy(self, tested_problems, f):
"""
Write matrix of accuracy of similarity/dissimilarity detection
Parameters:
- tested_problems -- list of tested problems
i-th element of list is either True or False
indicating if solutions of i-th problem
were tested
- f -- file to write to
"""
f.write("Matrix of similarity/dissimilarity detection accuracy\n")
f.write("-----------------------------------------------------\n")
_s = " "
for _i, _tested in enumerate(tested_problems):
if not _tested: continue
_s += "{:5d} ".format(_i)
f.write(_s + "\n")
for _i, _tested_row in enumerate(tested_problems):
if not _tested_row: continue
_s = "{:3d} ".format(_i)
for _j, _tested_col in enumerate(tested_problems):
if not _tested_col: continue
if self.sim_num_samples[_i, _j]:
_acc = self.sim_num_correct[_i, _j] / self.sim_num_samples[_i, _j]
_s += "{:5.1f} ".format(100.0 * _acc)
else:
_s += " - "
f.write(_s + "\n")
def reportSimilarityProblems(self, sim_accuracy, f):
"""
Write down accuracy of similarity detection
of two solutions of same problem
Parameters:
- sim_accuracy -- list of accuracy of similarity detection
An element of the list is a pair:
<problem, accuracy>
- f -- file to write the report to
"""
f.write("# Probl Accu- N N Problem name\n")
f.write(" numb racy tests errors \n")
for _i, _p_acc in enumerate(sim_accuracy):
_p = _p_acc[0]
f.write("{:4d} {:4d} {:6.2f}% {:5d} {:5d} {:s}\n".
format(_i, _p, _p_acc[1] * 100.0,
self.sim_num_samples[_p, _p],
self.sim_num_samples[_p, _p] - self.sim_num_correct[_p, _p],
self.problems[_p]))
def getSimilarityMistakes(self, problems):
"""
Get list of errors in similarity detection
of problem solutions
Parameters:
- problems -- indices of problems to report
Returns: list of lists of indices of samples for which
it was incorrectly decided that they give dissimilar solutions
"""
#Number of mistakes to detect that two solutions
#of same problem are similar
_mistakes = [[] for i in range(len(problems))]
for _i in range(self._n_samples):
_problem1, _solution1, _problem2, _solution2 = self.annotations[_i]
if _problem1 != _problem2: continue
if (self.predictions[_i] != self.labels[_i] and
_problem1 in problems):
_bad_idx = problems.index(_problem1)
_mistakes[_bad_idx].append(_i)
return _mistakes
def reportSimilarityMisclass(self, sim_accuracy, mistakes, f):
"""
Report errors in similarity detection of problem solutions
Parameters:
- sim_accuracy -- list of accuracy of similarity detection
An element of the list is a pair:
<problem, accuracy>
- problems -- problems to report
- mistakes -- list of lists of indices of samples for which
it was incorrectly decided that they give
disssimilar solutions
- f -- to write report to
"""
for _i, _c in enumerate(sim_accuracy):
_p, _a = _c
if not mistakes[_i]: continue
f.write(f"#{_i} For problem {self.problems[_p]} number {_p} " +
f"having {100.0 * _a : .2f}% accuracy of similarity detection\n")
f.write(f"there are {len(mistakes[_i])} " +
"incorrectly detected missimilarities:\n")
self.writeMistakes(mistakes[_i], f)
def writeMistakes(self, mistakes, f):
"""
Report problem solutions whose similarity was
incorrectly classified
Parameters:
- mistakes -- list of indices of incorrectly
classified samples
- f -- to write report to
"""
def _getConf(s):
"""
Get confedence of predicting sample s
Convert sigmoid output to probability
"""
_p = self.probabilities[s][0]
return _p if self.predictions[s] else 1 - _p
_n_errs_report = min(len(mistakes), self.max_n_cases_report)
_line = ""
_punct = ", "
for _i, _err_sample in enumerate(sorted(mistakes,
key =_getConf, reverse = True)[: _n_errs_report]):
_p1, _solution1, _p2, _solution2 = \
self.annotations[_err_sample]
#Convert sigmoid output to probability
#_p = self.probabilities[_err_sample][0]
#_conf = _p if self.predictions[_err_sample] else 1 - _p
_conf = "{:6.2f}".format(100.0 * _getConf(_err_sample))
if len(_line) > 80:
f.write(_line + "\n")
_line = ""
if _i + 1 == len(mistakes):
_punct = ""
_line += (self.solutions[_p1][_solution1] +
"/" + self.solutions[_p2][_solution2] + '(' +
_conf + '%)' + _punct)
if _line: f.write(_line + "\n\n")
def reportSimilarity(self, sim_accuracy):
"""
Analyze and report information on detection of similarity
of problem solutions
Parameters:
- sim_accuracy -- list of accuracy of similarity detection
An element of the list is a pair: <problem, accuracy>
"""
with open(f"{self.report_dir}/{self.simil_stat_fn}", 'w') as _f:
_f.write("Statistics on similarity detection\n")
_f.write("----------------------------------\n")
self.reportSimilarityProblems(sim_accuracy, _f)
_sim_acc =\
sorted(sim_accuracy, key = lambda _s : _s[1])
_worst_problems = _sim_acc[: self._n_cases_report]
_best_problems = _sim_acc[-self._n_cases_report :]
_problems = list(zip(*(_worst_problems + _best_problems)))[0]
_mistakes = self.getSimilarityMistakes(_problems)
_f.write("\nProblems with worst similarity detection\n")
_f.write("------------------------------------------\n")
self.reportSimilarityProblems(_worst_problems, _f)
_f.write("\nSimilarity detection errors for worst problems:\n")
_f.write("-------------------------------------------------\n")
self.reportSimilarityMisclass(_worst_problems,
_mistakes[: len(_worst_problems)], _f)
_f.write("\nProblems with best similarity detection\n")
_f.write("------------------------------------------\n")
self.reportSimilarityProblems(_best_problems, _f)
_f.write("\nSimilarity detection errors for best problems:\n")
_f.write("-----------------------------------------------\n")
self.reportSimilarityMisclass(_best_problems,
_mistakes[len(_worst_problems) :], _f)
def reportDissimilarityProblems(self, tests, f):
"""
Print report on accuracy of dissimilarity detection
of solutions of pairs of problems
Parameters:
- tests -- list of accuracy of dissimilarity tests
Elements of this list are tuples
<problem1, problem2, acc1, acc2, acc1+acc2, abs(acc1-acc2)>
If any accuracy acc1(acc2) is not defined:
acc1(acc2) = None
acc1+acc2 = 2*acc2(2*acc1)
abs(acc1-acc2) = 0
* acc1 is for similarity of problem1 to problem2 solutions
* acc2 is for similarity of problem2 to problem1 solutions
- f -- File to write to
"""
f.write("# 1-st problem 2-nd problem \n")
f.write(" Indx Name tsts errs Acc-cy Indx Name Tsts Errs Acc-cy\n")
f.write("-----------------------------------------------------------------------------------\n")
for _i, _t in enumerate(tests):
_p1, _p2, _a1, _a2, _a12, _diff = _t
_acc1 = "{:6.2f}%".format(100.0 * _a1) if _a1 is not None else " - "
_acc2 = "{:6.2f}%".format(100.0 * _a2) if _a2 is not None else " - "
f.write("{:3d} {:4d} {:16s} {:4d} {:4d} {:5s} {:4d} {:16s} {:4d} {:4d} {:5s}\n".
format(_i, _p1, self.problems[_p1],
self.sim_num_samples[_p1, _p2],
self.sim_num_samples[_p1, _p2] - self.sim_num_correct[_p1, _p2],
_acc1,
_p2, self.problems[_p2],
self.sim_num_samples[_p2, _p1],
self.sim_num_samples[_p2, _p1] - self.sim_num_correct[_p2, _p1],
_acc2))
def getDissimilarityMistakes(self, tests):
"""
Get list of errors in dissimilarity detection
of problem solutions
Parameters:
- tests -- list of accuracy of dissimilarity tests
Elements of this list are tuples
<problem1, problem2, acc1, acc2, acc1+acc2, abs(acc1-acc2)>
If any accuracy acc1(acc2) is not defined:
acc1(acc2) = None
acc1+acc2 = 2*acc2(2*acc1)
abs(acc1-acc2) = 0
Returns: list of lists of indices of samples for which
it was incorrectly decided that they represent similar solutions
"""
#Number of mistakes to detect that two solutions
#of same problem are similar
_test_ids_1 = [(_x[0], _x[1]) for _x in tests]
_test_ids_2 = [(_x[1], _x[0]) for _x in tests]
_mistakes = [[] for i in range(len(tests))]
for _i in range(self._n_samples):
_problem1, _solution1, _problem2, _solution2 = self.annotations[_i]
if _problem1 == _problem2: continue
if self.predictions[_i] == self.labels[_i]: continue
_t = (_problem1, _problem2)
if _t in _test_ids_1:
_bad_idx = _test_ids_1.index(_t)
_mistakes[_bad_idx].append(_i)
if _t in _test_ids_2:
_bad_idx = _test_ids_2.index(_t)
_mistakes[_bad_idx].append(_i)
return _mistakes
def reportDissimilarityMisclass(self, tests, mistakes, f):
"""
Report errors in dissimilarity detection of problem solutions
Parameters:
- tests -- tests to report
list of accuracy of dissimilarity tests
Elements of this list are tuples
<problem1, problem2, acc1, acc2, acc1+acc2, abs(acc1-acc2)>
If any accuracy acc1(acc2) is not defined:
acc1(acc2) = None
acc1+acc2 = 2*acc2(2*acc1)
abs(acc1-acc2) = 0
- mistakes -- list of lists of indices of samples for which
- f -- file to write to
it was incorrectly decided that they give similar solutions
"""
for _i, _t in enumerate(tests):
if not mistakes[_i]: continue
_p1, _p2, _a1, _a2, _a12, _diff = _t
_acc1 = "{:6.2f}%".format(100.0 * _a1) if _a1 is not None else "-"
_acc2 = "{:6.2f}%".format(100.0 * _a2) if _a2 is not None else "-"
f.write(f"#{_i} For problem {self.problems[_p1]} number {_p1} " +
f"and problem {self.problems[_p2]} number {_p2} " +
f"having {_acc1} and {_acc2} " +
"accuracies of dissimilarity detection\n")
f.write(f"There are {len(mistakes[_i])} " +
"incorrectly detected similarities:\n")
self.writeMistakes(mistakes[_i], f)
def reportDissimilarity(self, dissim_acc):
"""
Analyze and report information on detection of dissimilarity
of problem solutions
Parameters:
- dissim_acc -- list of accuracy of dissimilarity tests
Elements of this list are tuples
<problem1, problem2, acc1, acc2, acc1+acc2, abs(acc1-acc2)>
If any accuracy acc1(acc2) is not defined:
acc1(acc2) = None
acc1+acc2 = 2*acc2(2*acc1)
abs(acc1-acc2) = 0
"""
#Sort according to accuracy minimum
dissim_acc.sort(key = lambda _s : _s[4])
_worst_tests = dissim_acc[: self._n_cases_report]
_best_tests = dissim_acc[-self._n_cases_report :]
_mistakes = self.getDissimilarityMistakes(_worst_tests + _best_tests)
with open(f"{self.report_dir}/{self.dissimil_stat_fn}", 'w') as _f:
_f.write("\nWorst dissimilarity tests:\n")
_f.write("----------------------------\n")
self.reportDissimilarityProblems(_worst_tests, _f)
self.reportDissimilarityMisclass(_worst_tests,
_mistakes[: len(_worst_tests)], _f)
_f.write("\nBest dissimilarity tests:\n")
_f.write("-------------------------\n")
self.reportDissimilarityProblems(_best_tests, _f)
self.reportDissimilarityMisclass(_best_tests,
_mistakes[len(_worst_tests) :], _f)
_f.write("\nTests with high asymmetry of similarity detection:\n")
_f.write("---------------------------------------------------\n")
#Sort according to accuracy discrepancy
dissim_acc.sort(key = lambda _s : _s[5])
_asymm_tests = dissim_acc[-self._n_cases_report :]
_mistakes = self.getDissimilarityMistakes(_asymm_tests)
self.reportDissimilarityProblems(_asymm_tests, _f)
self.reportDissimilarityMisclass(_asymm_tests, _mistakes, _f)
def writeReport(self):
"""
Write report on similarity analysis
"""
self.printConfusionReport()
self.similarityConfusion()
_tested_problems, _sim_acc, _dissim_acc = \
self.compSimTestAccuracy()
with open(f"{self.report_dir}/{self._conf_mat_fn}", 'w') as _f:
self.writeSimDissimAccuracy(_tested_problems, _f)
self.reportSimilarity(_sim_acc)
self.reportDissimilarity(_dissim_acc)
#---------------- End of class ConfusionAnalysis ----------------------
|