# save_weights_only=False, mode='auto', period=1)
# tb = TensorBoard(log_dir=util.SUMMARY_DIR, histogram_freq=0.2, batch_size=batch_size, write_graph=True,
# write_grads=False, write_images=False, embeddings_freq=0, embeddings_layer_names=None,
# embeddings_metadata=None)
# model.fit(X_train, Y_train, batch_size=batch_size,
# epochs=100, verbose=0, callbacks=[csv_logger, tb, checkpoint], validation_data=(X_test, Y_test))
# print(X_test,Y_test)
# score = model.evaluate(X_test, Y_test, verbose=1, batch_size=batch_size)
# print(score)
model_name = 'LR_TL_BI_cuDnnGRU_N196_GloVe'
util = load_data(model_name, data_source='../Sentiment Analysis Data/Collected Dataset/'
'WorldCupDraw Preprocessed Dataset wo Dups.csv', data_label='text',
target_label='positive_percentage')
X = util.X
Y = util.Y
batch_size = 256
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.20, random_state=42)
validation_size = int(len(X_test) * 0.20)
model = gru_model(util)
# model = load_model('models/LR_TL_BI_cuDnnLSTM_N196_GloVe_e20/10-59.01-134.28.hdf5')
csv_logger = CSVLogger(util.LOG_DIR, append=True)
checkpoint_name = util.MODEL_DIR + '/{epoch:02d}-{loss:.2f}-{val_loss:.2f}.hdf5'
checkpoint = ModelCheckpoint(checkpoint_name, monitor='val_loss',
verbose=0, save_best_only=True,
save_weights_only=False, mode='auto', period=1)
tb = TensorBoard(log_dir=util.SUMMARY_DIR, histogram_freq=0.2, batch_size=batch_size, write_graph=True,
from keras.callbacks import TensorBoard, CSVLogger, ModelCheckpoint
from keras.layers import Dense, Embedding, CuDNNLSTM, Bidirectional
from keras.models import Sequential
from sklearn.model_selection import StratifiedKFold
from pandas import get_dummies
from Embedding_Utilities import load_data
util = load_data('TL_BI_cuDNNLSTM_L2_N196_GloVe_5f_final')
lstm_out = 196
model = Sequential()
model.add(
Embedding(util.NB_WORDS,
util.EMBEDDING_DIM,
input_length=util.TWEET_LENGTH,
dropout=0.2,
weights=[util.embedding_matrix],
trainable=False))
model.add(Bidirectional(CuDNNLSTM(lstm_out, return_sequences=True)))
model.add(Bidirectional(CuDNNLSTM(lstm_out, go_backwards=True)))
model.add(Dense(util.NUM_CLASSES, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='nadam',
metrics=['accuracy'])
print(model.summary())
kfold = StratifiedKFold(n_splits=5,
shuffle=True,
random_state=util.RANDOM_SEED)
cvscores = []
#
# result = model.predict(X[x].reshape(1, X.shape[1]), batch_size=1, verbose=2)[0]
#
# if argmax(result) == argmax(Y[x]):
# if argmax(Y[x]) == 0:
# neg_correct += 1
# else:
# pos_correct += 1
#
# if argmax(Y[x]) == 0:
# neg_cnt += 1
# else:
# pos_cnt += 1
#
# print("pos_acc", pos_correct / pos_cnt * 100, "%")
# print("neg_acc", neg_correct / neg_cnt * 100, "%")
model_to_load = "models/LR_TL_BI_cuDnnLSTM_N196_GloVe_e100/69-25.80-124.41.hdf5"
util = load_data('LR_TL_BI_cuDnnLSTM_N196_GloVe_e100',
data_source='../Sentiment Analysis Data/Collected Dataset/'
'WorldCupDraw Preprocessed Dataset wo Dups.csv',
data_label='text',
target_label='positive_percentage')
X = asarray([
'Germany vs. Mexico will be fun!',
'The moment England were drawn with Belgium... #bbcfootball'
])
X = util.convert_data(X)
model = models.load_model(model_to_load)
print(model.predict(X))
from keras import models
from Embedding_Utilities import load_data
from os.path import join
import pandas as pd
from numpy import argmax
model_name = "TL_BI_cuDNNLSTM_L2_N196_GloVe_5f_final/5.02-0.24-0.27.hdf5"
model_path = join('models/', model_name)
util = load_data(model_name)
replies_data = pd.read_csv(
'../Sentiment Analysis Data/Collected Dataset/Replies_WorldCupDraw_Cleansed.csv'
)
replies_tweet = util.convert_data(replies_data.text.values)
model = models.load_model(model_path)
batch_size = 256
result = model.predict(replies_tweet, batch_size=batch_size, verbose=1)
prediction = []
total_count = replies_data.text.size
neg_cnt, pos_cnt = 0, 0
for tweet_number in range(replies_data.text.size):
prediction.append(argmax(result[tweet_number]))
if argmax(result[tweet_number]) == 0:
neg_cnt += 1
else: