def test_CheckIfSameLength(self):
dbinfo = GetDatabaseInfoScholarshipsWithClassStatuses('Junior')
descriptionsList = dbinfo.getScholarshipDescriptionsList()
eligibilitiesList = dbinfo.getEligibilitiesList()
self.assertIsNotNone(descriptionsList)
self.assertIsNotNone(eligibilitiesList)
self.assertEqual(len(descriptionsList), len(eligibilitiesList))
def test_eligibilitiesList(self):
dbinfo = GetDatabaseInfoScholarshipsWithClassStatuses('Senior')
self.assertIsNotNone(dbinfo)
eligibilitesList = dbinfo.getEligibilitiesList()
self.assertIsNotNone(eligibilitesList)
testEligibility = eligibilitesList[0]
testCleanText = CleanText.cleanALLtheText(testEligibility)
self.assertIsNotNone(testCleanText)
def test_scholarshipsDescriptionsList(self):
dbinfo = GetDatabaseInfoScholarshipsWithClassStatuses('Junior')
self.assertIsNotNone(dbinfo)
descriptionsList = dbinfo.getScholarshipDescriptionsList()
self.assertIsNotNone(descriptionsList)
testDescription = descriptionsList[0]
testCleanText = CleanText.cleanALLtheText(testDescription)
self.assertIsNotNone(testCleanText)
def test_runEnsembleClassifier(self):
databaseInfo = GetDatabaseInfoScholarshipsWithClassStatuses()
dataTextList = databaseInfo.getConcatenatedDescriptionsEligibilities()
idsList = databaseInfo.getScholarshipsWithClassStatusIdsList()
ensembleClassifyTest = LogisticRegressionMultilabelClassifyClassStatusFromPretrainedModels(dataTextList,
idsList)
ensembleClassifyTest.displayResults()
def test_OVRClassifier(self):
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getConcatenatedDescriptionsEligibilities()
labelsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getRequirementNeededList()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getScholarshipsWithClassStatusIdsList()
testClassify = OneVsRestClassifyPreviouslyUntrained(
dataTextList, labelsList, idsList, trainingPercentage=0.8)
testClassify.trainTestAndDisplayResults()
def test_makeTrainingTestingSetBadPercentage(self):
desiredClassStatus = 'Senior'
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=desiredClassStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=desiredClassStatus, useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
trainingPercentage = 2
trainingSetAndTestingSet = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=desiredClassStatus,
secondLabel='Other',
firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList,
firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds,
trainingPercentage=trainingPercentage)
self.assertIsNone(trainingSetAndTestingSet)
def __init__(self):
databaseInfo = GetDatabaseInfoScholarshipsWithClassStatuses()
self.actualLabels = databaseInfo.getRequirementNeededList()
self.predictedLabels = databaseInfo.getEnsembleClassifierPredictions()
self.actualLabels = self.convertActualLabelStringsToList()
self.predictedLabels = self.convertPredictedLabelStringsToList()
self.calculateExactMatchAccuracy()
self.calculateNumLabelsAccuracy()
print("Average Accuracy Across Labels: %.2f percent" %
self.calculateAverageAccuracyWithinLists(self.actualLabels,
self.predictedLabels))
self.calculateListAccuracyWithinMatchedGreaterOrLesserListLengths()
def test_makeTrainingTestingSets(self):
desiredClassStatus = 'Junior'
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=desiredClassStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=desiredClassStatus, useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
trainingPercentage = 0.9
trainingSetAndTestingSet = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=desiredClassStatus,
secondLabel='Other',
firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList,
firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds,
trainingPercentage=trainingPercentage)
training = trainingSetAndTestingSet[0]
testing = trainingSetAndTestingSet[1]
totalSetSize = len(training) + len(testing)
self.assertEqual(math.ceil(totalSetSize * trainingPercentage),
len(training))
self.assertEqual((len(goodClassStatusDataTextList) +
len(badClassStatusDataTextList)), totalSetSize)
def test_makeMultilabelTrainingAndTestingSets(self):
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getConcatenatedDescriptionsEligibilities()
labelsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getRequirementNeededList()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getScholarshipsWithClassStatusIdsList()
trainingPercentage = 0.9
trainingSet, testingSet = MakeDataSet.makeMultilabelTrainingAndTestingSet(
dataTextList, labelsList, idsList, trainingPercentage)
totalSetSize = len(trainingSet) + len(testingSet)
self.assertEqual(math.ceil(totalSetSize * trainingPercentage),
len(trainingSet))
def test_makeSureTheLabelsListIsAListOfListsOfStrings(self):
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getConcatenatedDescriptionsEligibilities()
labelsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getRequirementNeededList()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getScholarshipsWithClassStatusIdsList()
testClassify = OneVsRestClassifyPreviouslyUntrained(
dataTextList, labelsList, idsList, trainingPercentage=0.8)
trainingSet = testClassify.trainingSet
trainingLabels = [training['label'] for training in trainingSet]
self.assertEqual(type(trainingLabels), list)
self.assertEqual(type(trainingLabels[1]), list)
self.assertEqual(type(trainingLabels[1][0]), str)
def test_CheckLabelsListFormat(self):
# get the list of labels, 100
dbinfo = GetDatabaseInfoScholarshipsWithClassStatuses()
listOLabels = dbinfo.getRequirementNeededList()
self.assertEqual(type(listOLabels), list)
self.assertEqual(type(listOLabels[1]), list)
first100Labels = listOLabels[:100]
self.assertEqual(len(first100Labels), 100)
# get fake data
X, Y = make_multilabel_classification(n_classes=10, n_labels=3, allow_unlabeled=False)
# test the OVR
testClassifiier = OneVsRestClassifier(LogisticRegression())
testClassifiier.fit(X, first100Labels)
def test_SeeWhatTheMultilabelTrainingSetIsActuallyPassingAsLabels(self):
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getConcatenatedDescriptionsEligibilities()
labelsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getRequirementNeededList()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses(
).getScholarshipsWithClassStatusIdsList()
trainingPercentage = 0.9
trainingSet, testingSet = MakeDataSet.makeMultilabelTrainingAndTestingSet(
dataTextList, labelsList, idsList, trainingPercentage)
trainingLabels = [
trainingInstance['label'] for trainingInstance in trainingSet
]
self.assertEqual(type(trainingLabels[1]), list)
self.assertEqual(type(trainingLabels[1][0]), str)
def test_trainOVRLRModel(self):
modelSaveFile = 'OneVsRestLRTrainedClassifiers\OneVsRestLRTrainedModel'
featuresValueCountsSaveFile = 'OneVsRestLRTrainedClassifiers\OneVsRestLRTrainedFeaturesValueCounts'
# first check to see if can open files:
testOpenModelSaveFile = open(modelSaveFile, 'rb')
testOpenModelSaveFile.close()
testOpenFVCSaveFile = open(featuresValueCountsSaveFile, 'rb')
testOpenFVCSaveFile.close()
# do training and save to files
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses().getConcatenatedDescriptionsEligibilities()
labelsList = GetDatabaseInfoScholarshipsWithClassStatuses().getRequirementNeededList()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses().getScholarshipsWithClassStatusIdsList()
testClassify = OneVsRestClassifyPreviouslyUntrained(dataTextList, labelsList, idsList,
trainingPercentage=0.99)
testClassify.trainAndSaveOVRModel(modelSaveFile, featuresValueCountsSaveFile)
def test_RunEnsembleClassifierInsertResultsIntoDB(self):
# get data from db
dataTextList = GetDatabaseInfoScholarshipsWithClassStatuses().getConcatenatedDescriptionsEligibilities()
idsList = GetDatabaseInfoScholarshipsWithClassStatuses().getScholarshipsWithClassStatusIdsList()
# run classifier, return results
pretrainedOVRLRModelFilePath = '..\OneVsRestLRTrainedClassifiers\OneVsRestLRTrainedModel'
pretrainedOVRLRFeaturesValueCountsIndexesFilePath = '..\OneVsRestLRTrainedClassifiers\OneVsRestLRTrainedFeaturesValueCounts'
testClassifier = OneVsRestClassifyFromPretrainedModel(pretrainedOVRLRModelFilePath,
pretrainedOVRLRFeaturesValueCountsIndexesFilePath,
dataTextList, idsList)
predictions = testClassifier.getPredictions()
# insert results into db
db = SUDBConnect()
for scholarshipWithClassStatusID, prediction in zip(idsList, predictions):
prediction = ', '.join(prediction)
db.insertUpdateOrDeleteDB(
"update dbo.ScholarshipsWithClassStatuses set OneVsRestClassifierPrediction='" + prediction + "' where ScholarshipsWithClassStatusId = '" + str(
scholarshipWithClassStatusID) + "'")
def test_runModelOnFreshman(self):
classStatus = 'Freshman'
modelSaveFile = '..\ClassifierTrainedModels\%sClassStatusTrainedLRModel' % classStatus
featuresValueCountsSaveFile = '..\ClassifierTrainedModels\%sClassStatusTrainedFeaturesValueCounts' % classStatus
# test to make sure can open model and fvc save files:
testModelSaveOpen = open(modelSaveFile, 'rb')
testModelSaveOpen.close()
testFVCSaveOpen = open(featuresValueCountsSaveFile, 'rb')
testFVCSaveOpen.close()
# now do test
databaseInfo = GetDatabaseInfoScholarshipsWithClassStatuses()
dataTextList = databaseInfo.getConcatenatedDescriptionsEligibilities()
idsList = databaseInfo.getScholarshipsWithClassStatusIdsList()
testClassify = LogisticRegressionClassifyClassStatusFromPretrainedModel(trainedModelInputFile=modelSaveFile,
trainedFeaturesValueCountsIndexesFile=featuresValueCountsSaveFile,
testingDataTextList=dataTextList,
testingDataIdsList=idsList)
testClassify.displayResults()
def test_RunEnsembleClassifierInsertResultsIntoDB(self):
# get data from db
databaseInfoExtract = GetDatabaseInfoScholarshipsWithClassStatuses()
dataTextList = databaseInfoExtract.getConcatenatedDescriptionsEligibilities()
idsList = databaseInfoExtract.getScholarshipsWithClassStatusIdsList()
# run classifier, return results
ensembleClassifyTest = LogisticRegressionMultilabelClassifyClassStatusFromPretrainedModels(dataTextList,
idsList)
idsAndPredictionsList = ensembleClassifyTest.doAllClassificationsAndReturnFilteredPredictionsListsById()
# insert results into db
db = SUDBConnect()
ids = [prediction[0] for prediction in idsAndPredictionsList]
predictions = [prediction[1] for prediction in idsAndPredictionsList]
for scholarshipWithClassStatusID, prediction in zip(ids, predictions):
prediction = ', '.join(prediction)
db.insertUpdateOrDeleteDB(
"update dbo.ScholarshipsWithClassStatuses set EnsembleClassifierPrediction='" + prediction + "' where ScholarshipsWithClassStatusId = '" + str(
scholarshipWithClassStatusID) + "'")
def test_runModelOnFreshman(self):
classStatus = 'Freshman'
modelSaveFile = '..\ClassifierTrainedModels\%sClassStatusTrainedLRModel' % classStatus
featuresValueCountsSaveFile = '..\ClassifierTrainedModels\%sClassStatusTrainedFeaturesValueCounts' % classStatus
# test to make sure can open model and fvc save files:
testModelSaveOpen = open(modelSaveFile, 'rb')
testModelSaveOpen.close()
testFVCSaveOpen = open(featuresValueCountsSaveFile, 'rb')
testFVCSaveOpen.close()
# now do test
databaseInfo = GetDatabaseInfoScholarshipsWithClassStatuses()
dataTextList = databaseInfo.getConcatenatedDescriptionsEligibilities()
idsList = databaseInfo.getScholarshipsWithClassStatusIdsList()
testClassify = LogisticRegressionClassifyClassStatusFromPretrainedModel(
trainedModelInputFile=modelSaveFile,
trainedFeaturesValueCountsIndexesFile=featuresValueCountsSaveFile,
testingDataTextList=dataTextList,
testingDataIdsList=idsList)
testClassify.displayResults()
def __init__(self, classStatus, trainingPercentage, modelSaveFile=None, featuresValuesCountsSaveFile=None):
self.classStatus = classStatus
self.trainingPercentage = trainingPercentage
self.badLabel = 'Other'
self.modelSaveFile = modelSaveFile
self.featuresValueCountsSaveFile = featuresValuesCountsSaveFile
print('Creating datasets...')
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=self.classStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=self.classStatus, useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList()
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities()
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList()
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities()
trainingTestingList = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=self.classStatus, secondLabel=self.badLabel, firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList, firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds, trainingPercentage=0.8)
self.training = trainingTestingList[0]
self.testing = trainingTestingList[1]
self.dataFrame = self.makeDataFrame()
self.trainingVectors = []
self.testingVectors = []
self.featuresValueCountsIndexes = []
self.logisticRegressionClassifier = LogisticRegression()
def __init__(self,
classStatus,
trainingPercentage,
modelSaveFile=None,
featuresValuesCountsSaveFile=None):
self.classStatus = classStatus
self.trainingPercentage = trainingPercentage
self.badLabel = 'Other'
self.modelSaveFile = modelSaveFile
self.featuresValueCountsSaveFile = featuresValuesCountsSaveFile
print('Creating datasets...')
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=self.classStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(
requirementNeeded=self.classStatus, useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList(
)
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities(
)
trainingTestingList = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=self.classStatus,
secondLabel=self.badLabel,
firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList,
firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds,
trainingPercentage=0.8)
self.training = trainingTestingList[0]
self.testing = trainingTestingList[1]
self.dataFrame = self.makeDataFrame()
self.trainingVectors = []
self.testingVectors = []
self.featuresValueCountsIndexes = []
self.logisticRegressionClassifier = LogisticRegression()
def test_makeTrainingTestingSetBadPercentage(self):
desiredClassStatus = 'Senior'
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=desiredClassStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=desiredClassStatus,
useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList()
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities()
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList()
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities()
trainingPercentage = 2
trainingSetAndTestingSet = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=desiredClassStatus, secondLabel='Other', firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList, firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds, trainingPercentage=trainingPercentage)
self.assertIsNone(trainingSetAndTestingSet)
def test_makeTrainingTestingSets(self):
desiredClassStatus = 'Junior'
goodClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=desiredClassStatus)
badClassStatusDB = GetDatabaseInfoScholarshipsWithClassStatuses(requirementNeeded=desiredClassStatus,
useNot=True)
goodClassStatusIds = goodClassStatusDB.getScholarshipsWithClassStatusIdsList()
goodClassStatusDataTextList = goodClassStatusDB.getConcatenatedDescriptionsEligibilities()
badClassStatusIds = badClassStatusDB.getScholarshipsWithClassStatusIdsList()
badClassStatusDataTextList = badClassStatusDB.getConcatenatedDescriptionsEligibilities()
trainingPercentage = 0.9
trainingSetAndTestingSet = MakeDataSet.makeBinaryLabelTrainingAndTestingSet(
firstLabel=desiredClassStatus, secondLabel='Other', firstLabelTextList=goodClassStatusDataTextList,
secondLabelTextList=badClassStatusDataTextList, firstLabelIdsList=goodClassStatusIds,
secondLabelIdsList=badClassStatusIds, trainingPercentage=trainingPercentage)
training = trainingSetAndTestingSet[0]
testing = trainingSetAndTestingSet[1]
totalSetSize = len(training) + len(testing)
self.assertEqual(math.ceil(totalSetSize * trainingPercentage), len(training))
self.assertEqual((len(goodClassStatusDataTextList) + len(badClassStatusDataTextList)), totalSetSize)