def get_evaluation_report(groundTruth, predictResult):
acc = accuracy_score(groundTruth, predictResult)
pre = precision_score(groundTruth, predictResult)
rec = recall_score(groundTruth, predictResult)
f1 = f1_score(groundTruth, predictResult)
log_the_string('acc:%.2f,pre:%.2f,rec:%.2f,f1:%.2f' % (acc, pre, rec, f1))
return acc, pre, rec, f1
def save_dataset_2_h5(dataX, labelY, dataset_path='dataset.h5'):
log_the_string('now is saving to h5')
dataset_file = h5py.File(dataset_path, 'w')
dataset_file.create_dataset('dataX', data=dataX)
dataset_file.create_dataset('labelY', data=labelY)
dataset_file.close()
return 0
def use_gaussianNB(dataX, lableY):
modelPath = 'gnb.model'
if not os.path.isfile(modelPath):
log_the_string('use gaussianNB and train it save it...')
clf = GaussianNB()
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_nearesrNeighbors(dataX, lableY):
modelPath = 'nn.model'
if not os.path.isfile(modelPath):
log_the_string('use nearesrNeighbors and train it save it...')
clf = NearestCentroid()
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_SGD(dataX, lableY):
modelPath = 'sgd.model'
if not os.path.isfile(modelPath):
log_the_string('use SGD and train it save it...')
clf = SGDClassifier(loss="hinge", penalty="l2")
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_SVM(dataX, lableY):
modelPath = 'svm.model'
if not os.path.isfile(modelPath):
log_the_string('use SVM and train it save it...')
clf = LinearSVC()
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_decision_classify_tree(dataX, lableY):
modelPath = 'dt.model'
if not os.path.isfile(modelPath):
log_the_string('use decision tree and train it save it...')
clf = tree.DecisionTreeClassifier()
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_MLP(dataX, lableY):
modelPath = 'mlp.model'
if not os.path.isfile(modelPath):
log_the_string('use MLP and train it save it...')
clf = MLPClassifier(solver='lbfgs',
alpha=1e-5,
hidden_layer_sizes=(5, 2),
random_state=1)
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
def use_randomForest(dataX, lableY):
modelPath = 'randomForest.model'
if not os.path.isfile(modelPath):
log_the_string('use randomForest and train it save it...')
clf = RandomForestClassifier(n_estimators=10,
max_depth=None,
min_samples_split=2,
random_state=0)
clf = clf.fit(dataX, lableY)
joblib.dump(clf, modelPath)
else:
clf = joblib.load(modelPath)
return clf
dataX, labelY = load_dataset_from_h5() # dataX, labelY = batch_get_mfcc()
print('load dataX shape:', dataX.shape)
X_train, X_test, y_train, y_test = train_test_split(dataX,
labelY,
test_size=0.2)
dctModel = use_decision_classify_tree(X_train, y_train)
svmModel = use_SVM(X_train, y_train)
sgdModel = use_SGD(X_train, y_train)
nearModle = use_nearesrNeighbors(X_train, y_train)
gaussianNB_Model = use_gaussianNB(X_train, y_train)
randomForestModel = use_randomForest(X_train, y_train)
mlpModel = use_MLP(X_train, y_train)
dctPredict = dctModel.predict(X_test)
svmPredict = svmModel.predict(X_test)
sgdPredict = sgdModel.predict(X_test)
nearPredict = nearModle.predict(X_test)
gaussianNB_Predict = gaussianNB_Model.predict(X_test)
randomForestPredict = randomForestModel.predict(X_test)
mlpPredict = mlpModel.predict(X_test)
final_vote_res = dctPredict + svmPredict + sgdPredict + nearPredict + gaussianNB_Predict + randomForestPredict + mlpPredict
final_vote_res_0_1 = [1 if item > 3 else 0 for item in final_vote_res]
get_evaluation_report(y_test, dctPredict)
get_evaluation_report(y_test, svmPredict)
get_evaluation_report(y_test, final_vote_res_0_1)
end = time()
log_the_string('it takes %.2f s' % (end - start))
def load_dataset_from_h5(dataset_path='dataset.h5'):
log_the_string('now is loading from h5')
dataset_file = h5py.File(dataset_path, 'r')
dataX = dataset_file['dataX'][:]
labelY = dataset_file['labelY'][:]
return dataX, labelY
def get_mfcc(wavFilePath):
log_the_string('now is processing %s...' % wavFilePath)
sampleRate, audioData = wavread(wavFilePath)
audioData = numpy.array([item / 2**15 for item in audioData])
mfcc_feat = mfcc(audioData, sampleRate)
return mfcc_feat
