def __init__(self):
'''
Training parameters:
'''
self.batch_size=512
self.minLen=50
self.maxLen=200
self.maxWords=50000
self.max_features=self.maxWords+3
self.nb_epoch=5
self.trained=False
self.modelLoaded=False
self.dataPrepared=False
self.tokenizer = Tokenizer(maxWords=self.maxWords)
print('Build model...')
self.model = Sequential()
self.model.add(Embedding(self.max_features, 128))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=True))
self.model.add(Dropout(0.5))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=True))
self.model.add(Dropout(0.5))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=False))
self.model.add(Dense(input_dim=128, output_dim=1))
self.model.add(Activation('sigmoid'))
self.model.compile(loss='binary_crossentropy', optimizer='adam', class_mode="binary")
print('Model has been built!')
class GenderClassifier(object):
def __init__(self):
'''
Training parameters:
'''
self.batch_size=512
self.minLen=50
self.maxLen=200
self.maxWords=50000
self.max_features=self.maxWords+3
self.nb_epoch=5
self.trained=False
self.modelLoaded=False
self.dataPrepared=False
self.tokenizer = Tokenizer(maxWords=self.maxWords)
print('Build model...')
self.model = Sequential()
self.model.add(Embedding(self.max_features, 128))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=True))
self.model.add(Dropout(0.5))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=True))
self.model.add(Dropout(0.5))
self.model.add(LSTM(input_dim=128, output_dim=128,return_sequences=False))
self.model.add(Dense(input_dim=128, output_dim=1))
self.model.add(Activation('sigmoid'))
self.model.compile(loss='binary_crossentropy', optimizer='adam', class_mode="binary")
print('Model has been built!')
def load_data(self,dataSize=-1):
import pandas as pd
file_loc = os.path.dirname(os.path.realpath(__file__))
relative_path = "blogger_data_2.csv" # move dataset to examples directory
fullpath = os.path.join(file_loc, relative_path)
if dataSize==-1:
data=pd.read_csv(fullpath)
else:
data = pd.read_csv(fullpath, nrows=dataSize)
X = data['text'].values
X = [str(x) for x in X]
y=data['gender'].values
return X,y
def prepareData(self,evaluate=False):
print("Loading data...")
X,y = self.load_data() # Can increase up to 250K or so
print("Number of lines:", len(X))
print("Vectorizing sequence data...")
if not evaluate:
self.tokenizer.fit(X)
self.tokenizer.save("./data/Tokenizer.pkl")
X=self.tokenizer.transform(X)
_X=[]
_y=[]
for i in range(len(X)):
if self.minLen<=len(X[i]) and len(X[i])<=self.maxLen:
_X.append(X[i])
_y.append(y[i])
X=sequence.pad_sequences(_X, maxlen=self.maxLen)
y=_y
count_1=0
for i in range(len(y)):
if y[i]==1.0:
count_1+=1
print("female percentage:",(len(y)-count_1)/len(y))
print("male percentage:",count_1/len(y))
self.X_train, self.X_valid, self.y_train, self.y_valid = train_test_split(X, y, test_size=0.10, random_state=100)
self.dataPrepared=True
print("Data preparation completed!")
def evaluate(self):
'''
For debugging purpose
'''
if not (self.modelLoaded or self.trained):
print('Warning: Model data has not been loaded or the model has not been trained.')
print("Trying to load Model dada...")
self.loadModel()
if not self.dataPrepared:
self.prepareData(evaluate=True)
score, acc = self.model.evaluate(self.X_valid, self.y_valid, batch_size=self.batch_size, show_accuracy=True)
print("Evaluation results:", score, " ", acc)
def train(self):
if not self.dataPrepared:
self.prepareData()
checkpointer = ModelCheckpoint(filepath="./weights.hdf5", verbose=1, save_best_only=True)
stopper=EarlyStopping(monitor='val_loss', patience=50, verbose=0)
print("Training...")
self.model.fit(self.X_train, self.y_train, batch_size=self.batch_size, nb_epoch=self.nb_epoch, validation_data=(self.X_valid,self.y_valid), show_accuracy=True,callbacks=[checkpointer,stopper])
self.trained=True
print('Training completed!')
def loadModel(self):
print('Loading model...')
self.model.load_weights('./data/weights.hdf5')
self.tokenizer.load("./data/Tokenizer.pkl")
self.modelLoaded=True
print('Model loaded')
'''
text: list or string
return: vector of results
'''
def predict(self,text):
if not (self.modelLoaded or self.trained):
print('Warning: Model data has not been loaded or the model has not been trained.')
print("Trying to load Model dada...")
self.loadModel()
X=np.array(self.tokenizer.transform(text))
y=self.model.predict(X,batch_size=1)
return y
