def predictHeartDiseaseUsingML():
# use parser and find the user's query
jsonRequest = request.get_json()
data = [jsonRequest]
testDataFrame = pd.DataFrame(data)
#Perform the same preprocessing as performed on training dataset
testDataFrameEncoded = featureEncodingUsingOneHotEncoder(testDataFrame)
testDataFrameEncodedAndScaledDataset = featureScalingUsingNormalizer(
testDataFrameEncoded)
#Load classifier from file
xtest = testDataFrameEncodedAndScaledDataset.iloc[:, :-1].values
ytest = testDataFrameEncodedAndScaledDataset.iloc[:,
len(testDataFrameEncodedAndScaledDataset
.columns) - 1].values
labelencoder_ytest = LabelEncoder()
ytest = labelencoder_ytest.fit_transform(ytest)
# Predicting the Test set results
ytestpred = classifier.predict(xtest)
prediction = {'prediction': int(ytestpred[0])}
return jsonify(prediction)
def performPreprocessingBuildModelsAndEvaluateAccuracy(trainingDataSet, testingDataSet, arrayOfModels):
for i in range(1,len(arrayOfModels)):
print('***************************************************************************************************************************')
print('********************************************* Building Model-', i ,' As Below *************************************************')
print('\t -- Feature Selection: \t ', arrayOfModels[i][0], ' \n\t -- Feature Encoding: \t ', arrayOfModels[i][1], ' \n\t -- Feature Scaling: \t ', arrayOfModels[i][2], ' \n\t -- Classification: \t ', arrayOfModels[i][3], '\n')
trainingFileNameWithAbsolutePath, testingFileNameWithAbsolutePath = getPathToTrainingAndTestingDataSets()
trainingDataSet = loadCSV(trainingFileNameWithAbsolutePath)
testingDataSet = loadCSV(testingFileNameWithAbsolutePath)
labelName = getLabelName()
label = trainingDataSet[labelName]
#Combining the test and training datasets for preprocessing then together, because we observed that in sme datasets
#the values in the categorical columns in test dataset and train dataset are being different this causes issues while
#applying classification techniques
completeDataSet = pd.concat(( trainingDataSet, testingDataSet ))
#difficultyLevel = completeDataSet.pop('difficulty_level')
print("completeDataSet.shape: ",completeDataSet.shape)
print("completeDataSet.head: ",completeDataSet.head())
#Feature Selection
if arrayOfModels[i][0] == 'TheilsU':
#Perform feature selection using TheilU
completeDataSetAfterFeatuerSelection = featureSelectionUsingTheilU(completeDataSet)
elif arrayOfModels[i][0] == 'Chi-SquaredTest':
#Perform feature selection using Chi-squared Test
completeDataSetAfterFeatuerSelection = featureSelectionUsingChisquaredTest(completeDataSet)
elif arrayOfModels[i][0] == 'RandomForestClassifier':
#Perform feature selection using RandomForestClassifier
completeDataSetAfterFeatuerSelection = featureSelectionUsingRandomForestClassifier(completeDataSet)
elif arrayOfModels[i][0] == 'ExtraTreesClassifier':
#Perform feature selection using ExtraTreesClassifier
completeDataSetAfterFeatuerSelection = featureSelectionUsingExtraTreesClassifier(completeDataSet)
#Feature Encoding
if arrayOfModels[i][1] == 'LabelEncoder':
#Perform lable encoding to convert categorical values into label encoded features
completeEncodedDataSet = featureEncodingUsingLabelEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'OneHotEncoder':
#Perform OnHot encoding to convert categorical values into one-hot encoded features
completeEncodedDataSet = featureEncodingUsingOneHotEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'FrequencyEncoder':
#Perform Frequency encoding to convert categorical values into frequency encoded features
completeEncodedDataSet = featureEncodingUsingFrequencyEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'BinaryEncoder':
#Perform Binary encoding to convert categorical values into binary encoded features
completeEncodedDataSet = featureEncodingUsingBinaryEncoder(completeDataSetAfterFeatuerSelection)
#Feature Scaling
if arrayOfModels[i][2] == 'Min-Max':
#Perform MinMaxScaler to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingMinMaxScaler(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Binarizing':
#Perform Binarizing to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingBinarizer(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Normalizing':
#Perform Normalizing to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingNormalizer(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Standardization':
#Perform Standardization to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingStandardScalar(completeEncodedDataSet)
#Split the complete dataSet into training dataSet and testing dataSet
featuresInPreProcessedTrainingDataSet,featuresInPreProcessedTestingDataSet,labelInPreProcessedTrainingDataSet,labelInPreProcessedTestingDataSet = splitCompleteDataSetIntoTrainingSetAndTestingSet(completeEncodedAndScaledDataset)
trainingEncodedAndScaledDataset = pd.concat([featuresInPreProcessedTrainingDataSet, labelInPreProcessedTrainingDataSet], axis=1, sort=False)
testingEncodedAndScaledDataset = pd.concat([featuresInPreProcessedTestingDataSet, labelInPreProcessedTestingDataSet], axis=1, sort=False)
#Classification
if arrayOfModels[i][3] == 'DecisonTree':
#Perform classification using DecisionTreeClassifier
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingDecisionTreeClassifier(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'RandomForestClassifier':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingRandomForestClassifier(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'ExtraTreesClassifier':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingExtraTreesClassifier(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'LogisticRegressionRegression':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingLogisticRegression(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'LinearDiscriminantAnalysis':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingLinearDiscriminantAnalysis(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'GuassianNaiveBayes':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingGaussianNB(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
elif arrayOfModels[i][3] == 'KNN':
classifier, trainingAccuracyScore, testingAccuracyScore = classifyUsingKNNClassifier(trainingEncodedAndScaledDataset, testingEncodedAndScaledDataset)
arrayOfModels[i].append(trainingAccuracyScore)
arrayOfModels[i].append(testingAccuracyScore)
modelName = arrayOfModels[i][0]+"_"+arrayOfModels[i][1]+"_"+arrayOfModels[i][2]+"_"+arrayOfModels[i][3]
modelFileName = getPathToGenerateModels() + modelName+".pkl"
arrayOfModels[i].append(modelName)
arrayOfModels[i].append(modelFileName)
#Save the model to file
joblib.dump(classifier, modelFileName)
def performPreprocessing(trainingDataSet, testingDataSet, arrayOfModels):
for i in range(0,len(arrayOfModels)):
print('***************************************************************************************************************************')
print('********************************************* Building Model-', i ,' As Below *************************************************')
print('\t -- Feature Selection: \t ', arrayOfModels[i][0], ' \n\t -- Feature Encoding: \t ', arrayOfModels[i][1], ' \n\t -- Feature Scaling: \t ', arrayOfModels[i][2], '\n')
trainingFileNameWithAbsolutePath, testingFileNameWithAbsolutePath = getPathToTrainingAndTestingDataSets()
trainingDataSet = loadCSV(trainingFileNameWithAbsolutePath)
testingDataSet = loadCSV(testingFileNameWithAbsolutePath)
labelName = getLabelName()
label = trainingDataSet[labelName]
#Combining the test and training datasets for preprocessing then together, because we observed that in sme datasets
#the values in the categorical columns in test dataset and train dataset are being different this causes issues while
#applying classification techniques
completeDataSet = pd.concat(( trainingDataSet, testingDataSet ))
#difficultyLevel = completeDataSet.pop('difficulty_level')
print("completeDataSet.shape: ",completeDataSet.shape)
print("completeDataSet.head: ",completeDataSet.head())
#Feature Selection
if arrayOfModels[i][0] == 'TheilsU':
#Perform feature selection using TheilU
completeDataSetAfterFeatuerSelection = featureSelectionUsingTheilU(completeDataSet)
elif arrayOfModels[i][0] == 'Chi-SquaredTest':
#Perform feature selection using Chi-squared Test
completeDataSetAfterFeatuerSelection = featureSelectionUsingChisquaredTest(completeDataSet)
elif arrayOfModels[i][0] == 'RandomForestClassifier':
#Perform feature selection using RandomForestClassifier
completeDataSetAfterFeatuerSelection = featureSelectionUsingRandomForestClassifier(completeDataSet)
elif arrayOfModels[i][0] == 'ExtraTreesClassifier':
#Perform feature selection using ExtraTreesClassifier
completeDataSetAfterFeatuerSelection = featureSelectionUsingExtraTreesClassifier(completeDataSet)
#Feature Encoding
if arrayOfModels[i][1] == 'LabelEncoder':
#Perform lable encoding to convert categorical values into label encoded features
completeEncodedDataSet = featureEncodingUsingLabelEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'OneHotEncoder':
#Perform OnHot encoding to convert categorical values into one-hot encoded features
completeEncodedDataSet = featureEncodingUsingOneHotEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'FrequencyEncoder':
#Perform Frequency encoding to convert categorical values into frequency encoded features
completeEncodedDataSet = featureEncodingUsingFrequencyEncoder(completeDataSetAfterFeatuerSelection)
elif arrayOfModels[i][1] == 'BinaryEncoder':
#Perform Binary encoding to convert categorical values into binary encoded features
completeEncodedDataSet = featureEncodingUsingBinaryEncoder(completeDataSetAfterFeatuerSelection)
#Feature Scaling
if arrayOfModels[i][2] == 'Min-Max':
#Perform MinMaxScaler to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingMinMaxScaler(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Binarizing':
#Perform Binarizing to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingBinarizer(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Normalizing':
#Perform Normalizing to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingNormalizer(completeEncodedDataSet)
elif arrayOfModels[i][2] == 'Standardization':
#Perform Standardization to scale the features of the dataset into same range
completeEncodedAndScaledDataset = featureScalingUsingStandardScalar(completeEncodedDataSet)
#Split the complete dataSet into training dataSet and testing dataSet
featuresInPreProcessedTrainingDataSet,featuresInPreProcessedTestingDataSet,labelInPreProcessedTrainingDataSet,labelInPreProcessedTestingDataSet = splitCompleteDataSetIntoTrainingSetAndTestingSet(completeEncodedAndScaledDataset)
trainingEncodedAndScaledDataset = pd.concat([featuresInPreProcessedTrainingDataSet, labelInPreProcessedTrainingDataSet], axis=1, sort=False)
testingEncodedAndScaledDataset = pd.concat([featuresInPreProcessedTestingDataSet, labelInPreProcessedTestingDataSet], axis=1, sort=False)
return completeEncodedAndScaledDataset