Search moodle.org's
Developer Documentation
Developer Documentation
- Bug fixes for general core bugs in 3.10.x will end 8 November 2021 (12 months).
- Bug fixes for security issues in 3.10.x will end 9 May 2022 (18 months).
- PHP version: minimum PHP 7.2.0 Note: minimum PHP version has increased since Moodle 3.8. PHP 7.3.x and 7.4.x are supported too.
/analytics/tests/ -> prediction_test.php (source)
Differences Between: [Versions 310 and 311] [Versions 310 and 400] [Versions 310 and 401] [Versions 310 and 402] [Versions 310 and 403] [Versions 39 and 310]
1 <?php 2 // This file is part of Moodle - http://moodle.org/ 3 // 4 // Moodle is free software: you can redistribute it and/or modify 5 // it under the terms of the GNU General Public License as published by 6 // the Free Software Foundation, either version 3 of the License, or 7 // (at your option) any later version. 8 // 9 // Moodle is distributed in the hope that it will be useful, 10 // but WITHOUT ANY WARRANTY; without even the implied warranty of 11 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 // GNU General Public License for more details. 13 // 14 // You should have received a copy of the GNU General Public License 15 // along with Moodle. If not, see <http://www.gnu.org/licenses/>. 16 17 /** 18 * Unit tests for evaluation, training and prediction. 19 * 20 * NOTE: in order to execute this test using a separate server for the 21 * python ML backend you need to define these variables in your config.php file: 22 * 23 * define('TEST_MLBACKEND_PYTHON_HOST', '127.0.0.1'); 24 * define('TEST_MLBACKEND_PYTHON_PORT', 5000); 25 * define('TEST_MLBACKEND_PYTHON_USERNAME', 'default'); 26 * define('TEST_MLBACKEND_PYTHON_PASSWORD', 'sshhhh'); 27 * 28 * @package core_analytics 29 * @copyright 2017 David Monllaó {@link http://www.davidmonllao.com} 30 * @license http://www.gnu.org/copyleft/gpl.html GNU GPL v3 or later 31 */ 32 33 defined('MOODLE_INTERNAL') || die(); 34 35 global $CFG; 36 require_once (__DIR__ . '/fixtures/test_indicator_max.php'); 37 require_once (__DIR__ . '/fixtures/test_indicator_min.php'); 38 require_once (__DIR__ . '/fixtures/test_indicator_null.php'); 39 require_once (__DIR__ . '/fixtures/test_indicator_fullname.php'); 40 require_once (__DIR__ . '/fixtures/test_indicator_random.php'); 41 require_once (__DIR__ . '/fixtures/test_indicator_multiclass.php'); 42 require_once (__DIR__ . '/fixtures/test_target_shortname.php'); 43 require_once (__DIR__ . '/fixtures/test_target_shortname_multiclass.php'); 44 require_once (__DIR__ . '/fixtures/test_static_target_shortname.php'); 45 46 require_once (__DIR__ . '/../../course/lib.php'); 47 48 /** 49 * Unit tests for evaluation, training and prediction. 50 * 51 * @package core_analytics 52 * @copyright 2017 David Monllaó {@link http://www.davidmonllao.com} 53 * @license http://www.gnu.org/copyleft/gpl.html GNU GPL v3 or later 54 */ 55 class core_analytics_prediction_testcase extends advanced_testcase { 56 57 /** 58 * Purge all the mlbackend outputs. 59 * 60 * This is done automatically for mlbackends using the web server dataroot but 61 * other mlbackends may store files elsewhere and these files need to be removed. 62 * 63 * @return null 64 */ 65 public function tearDown(): void { 66 $this->setAdminUser(); 67 68 $models = \core_analytics\manager::get_all_models(); 69 foreach ($models as $model) { 70 $model->delete(); 71 } 72 } 73 74 /** 75 * test_static_prediction 76 * 77 * @return void 78 */ 79 public function test_static_prediction() { 80 global $DB; 81 82 $this->resetAfterTest(true); 83 $this->setAdminuser(); 84 85 $model = $this->add_perfect_model('test_static_target_shortname'); 86 $model->enable('\core\analytics\time_splitting\no_splitting'); 87 $this->assertEquals(1, $model->is_enabled()); 88 $this->assertEquals(1, $model->is_trained()); 89 90 // No training for static models. 91 $results = $model->train(); 92 $trainedsamples = $DB->get_records('analytics_train_samples', array('modelid' => $model->get_id())); 93 $this->assertEmpty($trainedsamples); 94 $this->assertEmpty($DB->count_records('analytics_used_files', 95 array('modelid' => $model->get_id(), 'action' => 'trained'))); 96 97 // Now we create 2 hidden courses (only hidden courses are getting predictions). 98 $courseparams = array('shortname' => 'aaaaaa', 'fullname' => 'aaaaaa', 'visible' => 0); 99 $course1 = $this->getDataGenerator()->create_course($courseparams); 100 $courseparams = array('shortname' => 'bbbbbb', 'fullname' => 'bbbbbb', 'visible' => 0); 101 $course2 = $this->getDataGenerator()->create_course($courseparams); 102 103 $result = $model->predict(); 104 105 // Var $course1 predictions should be 1 == 'a', $course2 predictions should be 0 == 'b'. 106 $correct = array($course1->id => 1, $course2->id => 0); 107 foreach ($result->predictions as $uniquesampleid => $predictiondata) { 108 list($sampleid, $rangeindex) = $model->get_time_splitting()->infer_sample_info($uniquesampleid); 109 110 // The range index is not important here, both ranges prediction will be the same. 111 $this->assertEquals($correct[$sampleid], $predictiondata->prediction); 112 } 113 114 // 1 range for each analysable. 115 $predictedranges = $DB->get_records('analytics_predict_samples', array('modelid' => $model->get_id())); 116 $this->assertCount(2, $predictedranges); 117 // 2 predictions for each range. 118 $this->assertEquals(2, $DB->count_records('analytics_predictions', 119 array('modelid' => $model->get_id()))); 120 121 // No new generated records as there are no new courses available. 122 $model->predict(); 123 $predictedranges = $DB->get_records('analytics_predict_samples', array('modelid' => $model->get_id())); 124 $this->assertCount(2, $predictedranges); 125 $this->assertEquals(2, $DB->count_records('analytics_predictions', 126 array('modelid' => $model->get_id()))); 127 } 128 129 /** 130 * test_model_contexts 131 */ 132 public function test_model_contexts() { 133 global $DB; 134 135 $this->resetAfterTest(true); 136 $this->setAdminuser(); 137 138 $misc = $DB->get_record('course_categories', ['name' => 'Miscellaneous']); 139 $miscctx = \context_coursecat::instance($misc->id); 140 141 $category = $this->getDataGenerator()->create_category(); 142 $categoryctx = \context_coursecat::instance($category->id); 143 144 // One course per category. 145 $courseparams = array('shortname' => 'aaaaaa', 'fullname' => 'aaaaaa', 'visible' => 0, 146 'category' => $category->id); 147 $course1 = $this->getDataGenerator()->create_course($courseparams); 148 $course1ctx = \context_course::instance($course1->id); 149 $courseparams = array('shortname' => 'bbbbbb', 'fullname' => 'bbbbbb', 'visible' => 0, 150 'category' => $misc->id); 151 $course2 = $this->getDataGenerator()->create_course($courseparams); 152 153 $model = $this->add_perfect_model('test_static_target_shortname'); 154 155 // Just 1 category. 156 $model->update(true, false, '\core\analytics\time_splitting\no_splitting', false, [$categoryctx->id]); 157 $this->assertCount(1, $model->predict()->predictions); 158 159 // Now with 2 categories. 160 $model->update(true, false, false, false, [$categoryctx->id, $miscctx->id]); 161 162 // The courses in the new category are processed. 163 $this->assertCount(1, $model->predict()->predictions); 164 165 // Clear the predictions generated by the model and predict() again. 166 $model->clear(); 167 $this->assertCount(2, $model->predict()->predictions); 168 169 // Course context restriction. 170 $model->update(true, false, '\core\analytics\time_splitting\no_splitting', false, [$course1ctx->id]); 171 172 // Nothing new as the course was already analysed. 173 $result = $model->predict(); 174 $this->assertTrue(empty($result->predictions)); 175 176 $model->clear(); 177 $this->assertCount(1, $model->predict()->predictions); 178 } 179 180 /** 181 * test_ml_training_and_prediction 182 * 183 * @dataProvider provider_ml_training_and_prediction 184 * @param string $timesplittingid 185 * @param int $predictedrangeindex 186 * @param int $nranges 187 * @param string $predictionsprocessorclass 188 * @param array $forcedconfig 189 * @return void 190 */ 191 public function test_ml_training_and_prediction($timesplittingid, $predictedrangeindex, $nranges, $predictionsprocessorclass, 192 $forcedconfig) { 193 global $DB; 194 195 $this->resetAfterTest(true); 196 197 $this->set_forced_config($forcedconfig); 198 $predictionsprocessor = $this->is_predictions_processor_ready($predictionsprocessorclass); 199 200 $this->setAdminuser(); 201 set_config('enabled_stores', 'logstore_standard', 'tool_log'); 202 203 // Generate training data. 204 $ncourses = 10; 205 $this->generate_courses($ncourses); 206 207 $model = $this->add_perfect_model(); 208 209 $model->update(true, false, $timesplittingid, get_class($predictionsprocessor)); 210 211 // No samples trained yet. 212 $this->assertEquals(0, $DB->count_records('analytics_train_samples', array('modelid' => $model->get_id()))); 213 214 $results = $model->train(); 215 $this->assertEquals(1, $model->is_enabled()); 216 $this->assertEquals(1, $model->is_trained()); 217 218 // 20 courses * the 3 model indicators * the number of time ranges of this time splitting method. 219 $indicatorcalc = 20 * 3 * $nranges; 220 $this->assertEquals($indicatorcalc, $DB->count_records('analytics_indicator_calc')); 221 222 // 1 training file was created. 223 $trainedsamples = $DB->get_records('analytics_train_samples', array('modelid' => $model->get_id())); 224 $this->assertCount(1, $trainedsamples); 225 $samples = json_decode(reset($trainedsamples)->sampleids, true); 226 $this->assertCount($ncourses * 2, $samples); 227 $this->assertEquals(1, $DB->count_records('analytics_used_files', 228 array('modelid' => $model->get_id(), 'action' => 'trained'))); 229 // Check that analysable files for training are stored under labelled filearea. 230 $fs = get_file_storage(); 231 $this->assertCount(1, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 232 \core_analytics\dataset_manager::LABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 233 $this->assertEmpty($fs->get_directory_files(\context_system::instance()->id, 'analytics', 234 \core_analytics\dataset_manager::UNLABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 235 236 $params = [ 237 'startdate' => mktime(0, 0, 0, 10, 24, 2015), 238 'enddate' => mktime(0, 0, 0, 2, 24, 2016), 239 ]; 240 $courseparams = $params + array('shortname' => 'aaaaaa', 'fullname' => 'aaaaaa', 'visible' => 0); 241 $course1 = $this->getDataGenerator()->create_course($courseparams); 242 $courseparams = $params + array('shortname' => 'bbbbbb', 'fullname' => 'bbbbbb', 'visible' => 0); 243 $course2 = $this->getDataGenerator()->create_course($courseparams); 244 245 // They will not be skipped for prediction though. 246 $result = $model->predict(); 247 248 // Var $course1 predictions should be 1 == 'a', $course2 predictions should be 0 == 'b'. 249 $correct = array($course1->id => 1, $course2->id => 0); 250 foreach ($result->predictions as $uniquesampleid => $predictiondata) { 251 list($sampleid, $rangeindex) = $model->get_time_splitting()->infer_sample_info($uniquesampleid); 252 253 // The range index is not important here, both ranges prediction will be the same. 254 $this->assertEquals($correct[$sampleid], $predictiondata->prediction); 255 } 256 257 // 1 range will be predicted. 258 $predictedranges = $DB->get_records('analytics_predict_samples', array('modelid' => $model->get_id())); 259 $this->assertCount(1, $predictedranges); 260 foreach ($predictedranges as $predictedrange) { 261 $this->assertEquals($predictedrangeindex, $predictedrange->rangeindex); 262 $sampleids = json_decode($predictedrange->sampleids, true); 263 $this->assertCount(2, $sampleids); 264 $this->assertContains($course1->id, $sampleids); 265 $this->assertContains($course2->id, $sampleids); 266 } 267 $this->assertEquals(1, $DB->count_records('analytics_used_files', 268 array('modelid' => $model->get_id(), 'action' => 'predicted'))); 269 // 2 predictions. 270 $this->assertEquals(2, $DB->count_records('analytics_predictions', 271 array('modelid' => $model->get_id()))); 272 273 // Check that analysable files to get predictions are stored under unlabelled filearea. 274 $this->assertCount(1, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 275 \core_analytics\dataset_manager::LABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 276 $this->assertCount(1, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 277 \core_analytics\dataset_manager::UNLABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 278 279 // No new generated files nor records as there are no new courses available. 280 $model->predict(); 281 $predictedranges = $DB->get_records('analytics_predict_samples', array('modelid' => $model->get_id())); 282 $this->assertCount(1, $predictedranges); 283 foreach ($predictedranges as $predictedrange) { 284 $this->assertEquals($predictedrangeindex, $predictedrange->rangeindex); 285 } 286 $this->assertEquals(1, $DB->count_records('analytics_used_files', 287 array('modelid' => $model->get_id(), 'action' => 'predicted'))); 288 $this->assertEquals(2, $DB->count_records('analytics_predictions', 289 array('modelid' => $model->get_id()))); 290 291 // New samples that can be used for prediction. 292 $courseparams = $params + array('shortname' => 'cccccc', 'fullname' => 'cccccc', 'visible' => 0); 293 $course3 = $this->getDataGenerator()->create_course($courseparams); 294 $courseparams = $params + array('shortname' => 'dddddd', 'fullname' => 'dddddd', 'visible' => 0); 295 $course4 = $this->getDataGenerator()->create_course($courseparams); 296 297 $result = $model->predict(); 298 299 $predictedranges = $DB->get_records('analytics_predict_samples', array('modelid' => $model->get_id())); 300 $this->assertCount(1, $predictedranges); 301 foreach ($predictedranges as $predictedrange) { 302 $this->assertEquals($predictedrangeindex, $predictedrange->rangeindex); 303 $sampleids = json_decode($predictedrange->sampleids, true); 304 $this->assertCount(4, $sampleids); 305 $this->assertContains($course1->id, $sampleids); 306 $this->assertContains($course2->id, $sampleids); 307 $this->assertContains($course3->id, $sampleids); 308 $this->assertContains($course4->id, $sampleids); 309 } 310 $this->assertEquals(2, $DB->count_records('analytics_used_files', 311 array('modelid' => $model->get_id(), 'action' => 'predicted'))); 312 $this->assertEquals(4, $DB->count_records('analytics_predictions', 313 array('modelid' => $model->get_id()))); 314 $this->assertCount(1, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 315 \core_analytics\dataset_manager::LABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 316 $this->assertCount(2, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 317 \core_analytics\dataset_manager::UNLABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 318 319 // New visible course (for training). 320 $course5 = $this->getDataGenerator()->create_course(array('shortname' => 'aaa', 'fullname' => 'aa')); 321 $course6 = $this->getDataGenerator()->create_course(); 322 $result = $model->train(); 323 $this->assertEquals(2, $DB->count_records('analytics_used_files', 324 array('modelid' => $model->get_id(), 'action' => 'trained'))); 325 $this->assertCount(2, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 326 \core_analytics\dataset_manager::LABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 327 $this->assertCount(2, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 328 \core_analytics\dataset_manager::UNLABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 329 330 // Confirm that the files associated to the model are deleted on clear and on delete. The ML backend deletion 331 // processes will be triggered by these actions and any exception there would result in a failed test. 332 $model->clear(); 333 $this->assertEquals(0, $DB->count_records('analytics_used_files', 334 array('modelid' => $model->get_id(), 'action' => 'trained'))); 335 $this->assertCount(0, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 336 \core_analytics\dataset_manager::LABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 337 $this->assertCount(0, $fs->get_directory_files(\context_system::instance()->id, 'analytics', 338 \core_analytics\dataset_manager::UNLABELLED_FILEAREA, $model->get_id(), '/analysable/', true, false)); 339 $model->delete(); 340 341 set_config('enabled_stores', '', 'tool_log'); 342 get_log_manager(true); 343 } 344 345 /** 346 * provider_ml_training_and_prediction 347 * 348 * @return array 349 */ 350 public function provider_ml_training_and_prediction() { 351 $cases = array( 352 'no_splitting' => array('\core\analytics\time_splitting\no_splitting', 0, 1), 353 'quarters' => array('\core\analytics\time_splitting\quarters', 3, 4) 354 ); 355 356 // We need to test all system prediction processors. 357 return $this->add_prediction_processors($cases); 358 } 359 360 /** 361 * test_ml_export_import 362 * 363 * @param string $predictionsprocessorclass The class name 364 * @param array $forcedconfig 365 * @dataProvider provider_ml_processors 366 */ 367 public function test_ml_export_import($predictionsprocessorclass, $forcedconfig) { 368 $this->resetAfterTest(true); 369 370 $this->set_forced_config($forcedconfig); 371 $predictionsprocessor = $this->is_predictions_processor_ready($predictionsprocessorclass); 372 373 $this->setAdminuser(); 374 set_config('enabled_stores', 'logstore_standard', 'tool_log'); 375 376 // Generate training data. 377 $ncourses = 10; 378 $this->generate_courses($ncourses); 379 380 $model = $this->add_perfect_model(); 381 382 $model->update(true, false, '\core\analytics\time_splitting\quarters', get_class($predictionsprocessor)); 383 384 $model->train(); 385 $this->assertTrue($model->trained_locally()); 386 387 $this->generate_courses(10, ['visible' => 0]); 388 389 $originalresults = $model->predict(); 390 391 $zipfilename = 'model-zip-' . microtime() . '.zip'; 392 $zipfilepath = $model->export_model($zipfilename); 393 394 $modelconfig = new \core_analytics\model_config(); 395 list($modelconfig, $mlbackend) = $modelconfig->extract_import_contents($zipfilepath); 396 $this->assertNotFalse($mlbackend); 397 398 $importmodel = \core_analytics\model::import_model($zipfilepath); 399 $importmodel->enable(); 400 401 // Now predict using the imported model without prior training. 402 $importedmodelresults = $importmodel->predict(); 403 404 foreach ($originalresults->predictions as $sampleid => $prediction) { 405 $this->assertEquals($importedmodelresults->predictions[$sampleid]->prediction, $prediction->prediction); 406 } 407 408 $this->assertFalse($importmodel->trained_locally()); 409 410 $zipfilename = 'model-zip-' . microtime() . '.zip'; 411 $zipfilepath = $model->export_model($zipfilename, false); 412 413 $modelconfig = new \core_analytics\model_config(); 414 list($modelconfig, $mlbackend) = $modelconfig->extract_import_contents($zipfilepath); 415 $this->assertFalse($mlbackend); 416 417 set_config('enabled_stores', '', 'tool_log'); 418 get_log_manager(true); 419 } 420 421 /** 422 * provider_ml_processors 423 * 424 * @return array 425 */ 426 public function provider_ml_processors() { 427 $cases = [ 428 'case' => [], 429 ]; 430 431 // We need to test all system prediction processors. 432 return $this->add_prediction_processors($cases); 433 } 434 /** 435 * Test the system classifiers returns. 436 * 437 * This test checks that all mlbackend plugins in the system are able to return proper status codes 438 * even under weird situations. 439 * 440 * @dataProvider provider_ml_classifiers_return 441 * @param int $success 442 * @param int $nsamples 443 * @param int $classes 444 * @param string $predictionsprocessorclass 445 * @param array $forcedconfig 446 * @return void 447 */ 448 public function test_ml_classifiers_return($success, $nsamples, $classes, $predictionsprocessorclass, $forcedconfig) { 449 $this->resetAfterTest(); 450 451 $this->set_forced_config($forcedconfig); 452 $predictionsprocessor = $this->is_predictions_processor_ready($predictionsprocessorclass); 453 454 if ($nsamples % count($classes) != 0) { 455 throw new \coding_exception('The number of samples should be divisible by the number of classes'); 456 } 457 $samplesperclass = $nsamples / count($classes); 458 459 // Metadata (we pass 2 classes even if $classes only provides 1 class samples as we want to test 460 // what the backend does in this case. 461 $dataset = "nfeatures,targetclasses,targettype" . PHP_EOL; 462 $dataset .= "3,\"[0,1]\",\"discrete\"" . PHP_EOL; 463 464 // Headers. 465 $dataset .= "feature1,feature2,feature3,target" . PHP_EOL; 466 foreach ($classes as $class) { 467 for ($i = 0; $i < $samplesperclass; $i++) { 468 $dataset .= "1,0,1,$class" . PHP_EOL; 469 } 470 } 471 472 $trainingfile = array( 473 'contextid' => \context_system::instance()->id, 474 'component' => 'analytics', 475 'filearea' => 'labelled', 476 'itemid' => 123, 477 'filepath' => '/', 478 'filename' => 'whocares.csv' 479 ); 480 $fs = get_file_storage(); 481 $dataset = $fs->create_file_from_string($trainingfile, $dataset); 482 483 // Training should work correctly if at least 1 sample of each class is included. 484 $dir = make_request_directory(); 485 $modeluniqueid = 'whatever' . microtime(); 486 $result = $predictionsprocessor->train_classification($modeluniqueid, $dataset, $dir); 487 488 switch ($success) { 489 case 'yes': 490 $this->assertEquals(\core_analytics\model::OK, $result->status); 491 break; 492 case 'no': 493 $this->assertNotEquals(\core_analytics\model::OK, $result->status); 494 break; 495 case 'maybe': 496 default: 497 // We just check that an object is returned so we don't have an empty check, 498 // what we really want to check is that an exception was not thrown. 499 $this->assertInstanceOf(\stdClass::class, $result); 500 } 501 502 // Purge the directory used in this test (useful in case the mlbackend is storing files 503 // somewhere out of the default moodledata/models dir. 504 $predictionsprocessor->delete_output_dir($dir, $modeluniqueid); 505 } 506 507 /** 508 * test_ml_classifiers_return provider 509 * 510 * We can not be very specific here as test_ml_classifiers_return only checks that 511 * mlbackend plugins behave and expected and control properly backend errors even 512 * under weird situations. 513 * 514 * @return array 515 */ 516 public function provider_ml_classifiers_return() { 517 // Using verbose options as the first argument for readability. 518 $cases = array( 519 '1-samples' => array('maybe', 1, [0]), 520 '2-samples-same-class' => array('maybe', 2, [0]), 521 '2-samples-different-classes' => array('yes', 2, [0, 1]), 522 '4-samples-different-classes' => array('yes', 4, [0, 1]) 523 ); 524 525 // We need to test all system prediction processors. 526 return $this->add_prediction_processors($cases); 527 } 528 529 /** 530 * Tests correct multi-classification. 531 * 532 * @dataProvider provider_test_multi_classifier 533 * @param string $timesplittingid 534 * @param string $predictionsprocessorclass 535 * @param array|null $forcedconfig 536 * @throws coding_exception 537 * @throws moodle_exception 538 */ 539 public function test_ml_multi_classifier($timesplittingid, $predictionsprocessorclass, $forcedconfig) { 540 global $DB; 541 542 $this->resetAfterTest(true); 543 $this->setAdminuser(); 544 set_config('enabled_stores', 'logstore_standard', 'tool_log'); 545 546 $this->set_forced_config($forcedconfig); 547 548 $predictionsprocessor = \core_analytics\manager::get_predictions_processor($predictionsprocessorclass, false); 549 if ($predictionsprocessor->is_ready() !== true) { 550 $this->markTestSkipped('Skipping ' . $predictionsprocessorclass . ' as the predictor is not ready.'); 551 } 552 // Generate training courses. 553 $ncourses = 5; 554 $this->generate_courses_multiclass($ncourses); 555 $model = $this->add_multiclass_model(); 556 $model->update(true, false, $timesplittingid, get_class($predictionsprocessor)); 557 $results = $model->train(); 558 559 $params = [ 560 'startdate' => mktime(0, 0, 0, 10, 24, 2015), 561 'enddate' => mktime(0, 0, 0, 2, 24, 2016), 562 ]; 563 $courseparams = $params + array('shortname' => 'aaaaaa', 'fullname' => 'aaaaaa', 'visible' => 0); 564 $course1 = $this->getDataGenerator()->create_course($courseparams); 565 $courseparams = $params + array('shortname' => 'bbbbbb', 'fullname' => 'bbbbbb', 'visible' => 0); 566 $course2 = $this->getDataGenerator()->create_course($courseparams); 567 $courseparams = $params + array('shortname' => 'cccccc', 'fullname' => 'cccccc', 'visible' => 0); 568 $course3 = $this->getDataGenerator()->create_course($courseparams); 569 570 // They will not be skipped for prediction though. 571 $result = $model->predict(); 572 // The $course1 predictions should be 0 == 'a', $course2 should be 1 == 'b' and $course3 should be 2 == 'c'. 573 $correct = array($course1->id => 0, $course2->id => 1, $course3->id => 2); 574 foreach ($result->predictions as $uniquesampleid => $predictiondata) { 575 list($sampleid, $rangeindex) = $model->get_time_splitting()->infer_sample_info($uniquesampleid); 576 577 // The range index is not important here, both ranges prediction will be the same. 578 $this->assertEquals($correct[$sampleid], $predictiondata->prediction); 579 } 580 581 set_config('enabled_stores', '', 'tool_log'); 582 get_log_manager(true); 583 } 584 585 /** 586 * Provider for the multi_classification test. 587 * 588 * @return array 589 */ 590 public function provider_test_multi_classifier() { 591 $cases = array( 592 'notimesplitting' => array('\core\analytics\time_splitting\no_splitting'), 593 ); 594 595 // Add all system prediction processors. 596 return $this->add_prediction_processors($cases); 597 } 598 599 /** 600 * Basic test to check that prediction processors work as expected. 601 * 602 * @coversNothing 603 * @dataProvider provider_ml_test_evaluation_configuration 604 * @param string $modelquality 605 * @param int $ncourses 606 * @param array $expected 607 * @param string $predictionsprocessorclass 608 * @param array $forcedconfig 609 * @return void 610 */ 611 public function test_ml_evaluation_configuration($modelquality, $ncourses, $expected, $predictionsprocessorclass, 612 $forcedconfig) { 613 $this->resetAfterTest(true); 614 615 $this->set_forced_config($forcedconfig); 616 $predictionsprocessor = $this->is_predictions_processor_ready($predictionsprocessorclass); 617 618 $this->setAdminuser(); 619 set_config('enabled_stores', 'logstore_standard', 'tool_log'); 620 621 $sometimesplittings = '\core\analytics\time_splitting\single_range,' . 622 '\core\analytics\time_splitting\quarters'; 623 set_config('defaulttimesplittingsevaluation', $sometimesplittings, 'analytics'); 624 625 if ($modelquality === 'perfect') { 626 $model = $this->add_perfect_model(); 627 } else if ($modelquality === 'random') { 628 $model = $this->add_random_model(); 629 } else { 630 throw new \coding_exception('Only perfect and random accepted as $modelquality values'); 631 } 632 633 // Generate training data. 634 $this->generate_courses($ncourses); 635 636 $model->update(false, false, false, get_class($predictionsprocessor)); 637 $results = $model->evaluate(); 638 639 // We check that the returned status includes at least $expectedcode code. 640 foreach ($results as $timesplitting => $result) { 641 $message = 'The returned status code ' . $result->status . ' should include ' . $expected[$timesplitting]; 642 $filtered = $result->status & $expected[$timesplitting]; 643 $this->assertEquals($expected[$timesplitting], $filtered, $message); 644 645 $options = ['evaluation' => true, 'reuseprevanalysed' => true]; 646 $result = new \core_analytics\local\analysis\result_file($model->get_id(), true, $options); 647 $timesplittingobj = \core_analytics\manager::get_time_splitting($timesplitting); 648 $analysable = new \core_analytics\site(); 649 $cachedanalysis = $result->retrieve_cached_result($timesplittingobj, $analysable); 650 $this->assertInstanceOf(\stored_file::class, $cachedanalysis); 651 } 652 653 set_config('enabled_stores', '', 'tool_log'); 654 get_log_manager(true); 655 } 656 657 /** 658 * Tests the evaluation of already trained models. 659 * 660 * @coversNothing 661 * @dataProvider provider_ml_processors 662 * @param string $predictionsprocessorclass 663 * @param array $forcedconfig 664 * @return null 665 */ 666 public function test_ml_evaluation_trained_model($predictionsprocessorclass, $forcedconfig) { 667 $this->resetAfterTest(true); 668 669 $this->set_forced_config($forcedconfig); 670 $predictionsprocessor = $this->is_predictions_processor_ready($predictionsprocessorclass); 671 672 $this->setAdminuser(); 673 set_config('enabled_stores', 'logstore_standard', 'tool_log'); 674 675 $model = $this->add_perfect_model(); 676 677 // Generate training data. 678 $this->generate_courses(50); 679 680 $model->update(true, false, '\\core\\analytics\\time_splitting\\quarters', get_class($predictionsprocessor)); 681 $model->train(); 682 683 $zipfilename = 'model-zip-' . microtime() . '.zip'; 684 $zipfilepath = $model->export_model($zipfilename); 685 $importmodel = \core_analytics\model::import_model($zipfilepath); 686 687 $results = $importmodel->evaluate(['mode' => 'trainedmodel']); 688 $this->assertEquals(0, $results['\\core\\analytics\\time_splitting\\quarters']->status); 689 $this->assertEquals(1, $results['\\core\\analytics\\time_splitting\\quarters']->score); 690 691 set_config('enabled_stores', '', 'tool_log'); 692 get_log_manager(true); 693 } 694 695 /** 696 * test_read_indicator_calculations 697 * 698 * @return void 699 */ 700 public function test_read_indicator_calculations() { 701 global $DB; 702 703 $this->resetAfterTest(true); 704 705 $starttime = 123; 706 $endtime = 321; 707 $sampleorigin = 'whatever'; 708 709 $indicator = $this->getMockBuilder('test_indicator_max')->setMethods(['calculate_sample'])->getMock(); 710 $indicator->expects($this->never())->method('calculate_sample'); 711 712 $existingcalcs = array(111 => 1, 222 => -1); 713 $sampleids = array(111 => 111, 222 => 222); 714 list($values, $unused) = $indicator->calculate($sampleids, $sampleorigin, $starttime, $endtime, $existingcalcs); 715 } 716 717 /** 718 * test_not_null_samples 719 */ 720 public function test_not_null_samples() { 721 $this->resetAfterTest(true); 722 723 $timesplitting = \core_analytics\manager::get_time_splitting('\core\analytics\time_splitting\quarters'); 724 $timesplitting->set_analysable(new \core_analytics\site()); 725 726 $ranges = array( 727 array('start' => 111, 'end' => 222, 'time' => 222), 728 array('start' => 222, 'end' => 333, 'time' => 333) 729 ); 730 $samples = array(123 => 123, 321 => 321); 731 732 $target = \core_analytics\manager::get_target('test_target_shortname'); 733 $indicators = array('test_indicator_null', 'test_indicator_min'); 734 foreach ($indicators as $key => $indicator) { 735 $indicators[$key] = \core_analytics\manager::get_indicator($indicator); 736 } 737 $model = \core_analytics\model::create($target, $indicators, '\core\analytics\time_splitting\no_splitting'); 738 739 $analyser = $model->get_analyser(); 740 $result = new \core_analytics\local\analysis\result_array($model->get_id(), false, $analyser->get_options()); 741 $analysis = new \core_analytics\analysis($analyser, false, $result); 742 743 // Samples with at least 1 not null value are returned. 744 $params = array( 745 $timesplitting, 746 $samples, 747 $ranges 748 ); 749 $dataset = phpunit_util::call_internal_method($analysis, 'calculate_indicators', $params, 750 '\core_analytics\analysis'); 751 $this->assertArrayHasKey('123-0', $dataset); 752 $this->assertArrayHasKey('123-1', $dataset); 753 $this->assertArrayHasKey('321-0', $dataset); 754 $this->assertArrayHasKey('321-1', $dataset); 755 756 757 $indicators = array('test_indicator_null'); 758 foreach ($indicators as $key => $indicator) { 759 $indicators[$key] = \core_analytics\manager::get_indicator($indicator); 760 } 761 $model = \core_analytics\model::create($target, $indicators, '\core\analytics\time_splitting\no_splitting'); 762 763 $analyser = $model->get_analyser(); 764 $result = new \core_analytics\local\analysis\result_array($model->get_id(), false, $analyser->get_options()); 765 $analysis = new \core_analytics\analysis($analyser, false, $result); 766 767 // Samples with only null values are not returned. 768 $params = array( 769 $timesplitting, 770 $samples, 771 $ranges 772 ); 773 $dataset = phpunit_util::call_internal_method($analysis, 'calculate_indicators', $params, 774 '\core_analytics\analysis'); 775 $this->assertArrayNotHasKey('123-0', $dataset); 776 $this->assertArrayNotHasKey('123-1', $dataset); 777 $this->assertArrayNotHasKey('321-0', $dataset); 778 $this->assertArrayNotHasKey('321-1', $dataset); 779 } 780 781 /** 782 * provider_ml_test_evaluation_configuration 783 * 784 * @return array 785 */ 786 public function provider_ml_test_evaluation_configuration() { 787 788 $cases = array( 789 'bad' => array( 790 'modelquality' => 'random', 791 'ncourses' => 50, 792 'expectedresults' => array( 793 '\core\analytics\time_splitting\single_range' => \core_analytics\model::LOW_SCORE, 794 '\core\analytics\time_splitting\quarters' => \core_analytics\model::LOW_SCORE, 795 ) 796 ), 797 'good' => array( 798 'modelquality' => 'perfect', 799 'ncourses' => 50, 800 'expectedresults' => array( 801 '\core\analytics\time_splitting\single_range' => \core_analytics\model::OK, 802 '\core\analytics\time_splitting\quarters' => \core_analytics\model::OK, 803 ) 804 ) 805 ); 806 return $this->add_prediction_processors($cases); 807 } 808 809 /** 810 * add_random_model 811 * 812 * @return \core_analytics\model 813 */ 814 protected function add_random_model() { 815 816 $target = \core_analytics\manager::get_target('test_target_shortname'); 817 $indicators = array('test_indicator_max', 'test_indicator_min', 'test_indicator_random'); 818 foreach ($indicators as $key => $indicator) { 819 $indicators[$key] = \core_analytics\manager::get_indicator($indicator); 820 } 821 822 $model = \core_analytics\model::create($target, $indicators); 823 824 // To load db defaults as well. 825 return new \core_analytics\model($model->get_id()); 826 } 827 828 /** 829 * add_perfect_model 830 * 831 * @param string $targetclass 832 * @return \core_analytics\model 833 */ 834 protected function add_perfect_model($targetclass = 'test_target_shortname') { 835 $target = \core_analytics\manager::get_target($targetclass); 836 $indicators = array('test_indicator_max', 'test_indicator_min', 'test_indicator_fullname'); 837 foreach ($indicators as $key => $indicator) { 838 $indicators[$key] = \core_analytics\manager::get_indicator($indicator); 839 } 840 841 $model = \core_analytics\model::create($target, $indicators); 842 843 // To load db defaults as well. 844 return new \core_analytics\model($model->get_id()); 845 } 846 847 /** 848 * Generates model for multi-classification 849 * 850 * @param string $targetclass 851 * @return \core_analytics\model 852 * @throws coding_exception 853 * @throws moodle_exception 854 */ 855 public function add_multiclass_model($targetclass = 'test_target_shortname_multiclass') { 856 $target = \core_analytics\manager::get_target($targetclass); 857 $indicators = array('test_indicator_fullname', 'test_indicator_multiclass'); 858 foreach ($indicators as $key => $indicator) { 859 $indicators[$key] = \core_analytics\manager::get_indicator($indicator); 860 } 861 862 $model = \core_analytics\model::create($target, $indicators); 863 return new \core_analytics\model($model->get_id()); 864 } 865 866 /** 867 * Generates $ncourses courses 868 * 869 * @param int $ncourses The number of courses to be generated. 870 * @param array $params Course params 871 * @return null 872 */ 873 protected function generate_courses($ncourses, array $params = []) { 874 875 $params = $params + [ 876 'startdate' => mktime(0, 0, 0, 10, 24, 2015), 877 'enddate' => mktime(0, 0, 0, 2, 24, 2016), 878 ]; 879 880 for ($i = 0; $i < $ncourses; $i++) { 881 $name = 'a' . random_string(10); 882 $courseparams = array('shortname' => $name, 'fullname' => $name) + $params; 883 $this->getDataGenerator()->create_course($courseparams); 884 } 885 for ($i = 0; $i < $ncourses; $i++) { 886 $name = 'b' . random_string(10); 887 $courseparams = array('shortname' => $name, 'fullname' => $name) + $params; 888 $this->getDataGenerator()->create_course($courseparams); 889 } 890 } 891 892 /** 893 * Generates ncourses for multi-classification 894 * 895 * @param int $ncourses The number of courses to be generated. 896 * @param array $params Course params 897 * @return null 898 */ 899 protected function generate_courses_multiclass($ncourses, array $params = []) { 900 901 $params = $params + [ 902 'startdate' => mktime(0, 0, 0, 10, 24, 2015), 903 'enddate' => mktime(0, 0, 0, 2, 24, 2016), 904 ]; 905 906 for ($i = 0; $i < $ncourses; $i++) { 907 $name = 'a' . random_string(10); 908 $courseparams = array('shortname' => $name, 'fullname' => $name) + $params; 909 $this->getDataGenerator()->create_course($courseparams); 910 } 911 for ($i = 0; $i < $ncourses; $i++) { 912 $name = 'b' . random_string(10); 913 $courseparams = array('shortname' => $name, 'fullname' => $name) + $params; 914 $this->getDataGenerator()->create_course($courseparams); 915 } 916 for ($i = 0; $i < $ncourses; $i++) { 917 $name = 'c' . random_string(10); 918 $courseparams = array('shortname' => $name, 'fullname' => $name) + $params; 919 $this->getDataGenerator()->create_course($courseparams); 920 } 921 } 922 923 /** 924 * Forces some configuration values. 925 * 926 * @param array $forcedconfig 927 */ 928 protected function set_forced_config($forcedconfig) { 929 \core_analytics\manager::reset_prediction_processors(); 930 931 if (empty($forcedconfig)) { 932 return; 933 } 934 foreach ($forcedconfig as $pluginname => $pluginconfig) { 935 foreach ($pluginconfig as $name => $value) { 936 set_config($name, $value, $pluginname); 937 } 938 } 939 } 940 941 /** 942 * Is the provided processor ready using the current configuration in the site? 943 * 944 * @param string $predictionsprocessorclass 945 * @return \core_analytics\predictor 946 */ 947 protected function is_predictions_processor_ready(string $predictionsprocessorclass) { 948 // We repeat the test for all prediction processors. 949 $predictionsprocessor = \core_analytics\manager::get_predictions_processor($predictionsprocessorclass, false); 950 $ready = $predictionsprocessor->is_ready(); 951 if ($ready !== true) { 952 $this->markTestSkipped('Skipping ' . $predictionsprocessorclass . ' as the predictor is not ready: ' . $ready); 953 } 954 955 return $predictionsprocessor; 956 } 957 958 /** 959 * add_prediction_processors 960 * 961 * @param array $cases 962 * @return array 963 */ 964 protected function add_prediction_processors($cases) { 965 966 $return = array(); 967 968 if (defined('TEST_MLBACKEND_PYTHON_HOST') && defined('TEST_MLBACKEND_PYTHON_PORT') 969 && defined('TEST_MLBACKEND_PYTHON_USERNAME') && defined('TEST_MLBACKEND_PYTHON_USERNAME')) { 970 $testpythonserver = true; 971 } 972 973 // We need to test all prediction processors in the system. 974 $predictionprocessors = \core_analytics\manager::get_all_prediction_processors(); 975 foreach ($predictionprocessors as $classfullname => $predictionsprocessor) { 976 foreach ($cases as $key => $case) { 977 978 if (!$predictionsprocessor instanceof \mlbackend_python\processor || empty($testpythonserver)) { 979 $extraparams = ['predictionsprocessor' => $classfullname, 'forcedconfig' => null]; 980 $return[$key . '-' . $classfullname] = $case + $extraparams; 981 } else { 982 983 // We want the configuration to be forced during the test as things like importing models create new 984 // instances of ML backend processors during the process. 985 $forcedconfig = ['mlbackend_python' => ['useserver' => true, 'host' => TEST_MLBACKEND_PYTHON_HOST, 986 'port' => TEST_MLBACKEND_PYTHON_PORT, 'secure' => false, 'username' => TEST_MLBACKEND_PYTHON_USERNAME, 987 'password' => TEST_MLBACKEND_PYTHON_PASSWORD]]; 988 $casekey = $key . '-' . $classfullname . '-server'; 989 $return[$casekey] = $case + ['predictionsprocessor' => $classfullname, 'forcedconfig' => $forcedconfig]; 990 } 991 } 992 } 993 994 return $return; 995 } 996 }
title
Description
Body
title
Description
Body
title
Description
Body
title
Body
