def main():
global config
USE_CLASSIFIER='lstm'
weight_file_path = Config.getPath('models') + '/' + USE_CLASSIFIER + '-weights.h5'
load_config(USE_CLASSIFIER)
classifier=ClassifierFactory.getLSTM(**{'config':config})
#Classifier(model_name=USE_CLASSIFIER,config=config)
classifier.load_weights(weight_file_path)
df = pd.read_csv(Config.getPath('data') + '/' + TESTING_DATA)
Xtest = df['question_text']
Ytest = df['target']
print('extract configuration from input texts ...')
print('testing size: ', len(Xtest))
print('start predicting ...')
pred = classifier.predict(Xtest)
print(pred)
score = metrics.accuracy_score(Ytest, pred)
print("accuracy: %0.3f" % score)
cm = metrics.confusion_matrix(Ytest, pred, labels=[0, 1])
plot_confusion_matrix(cm, classes=[0, 1])
def load_config(model_name):
print('loading csv file ...')
global config
config_file_path = Config.getPath('models') + '/' + model_name + '-config.joblib'
config = ClassifierFactory.getConfig(joblib_file=config_file_path)
#Two classes - Fake=0, Reliable=1
config.set('num_target_tokens',2)
def predict_svm():
global config
load_config('svm')
print('loading data...')
df = pd.read_csv(Config.getPath('data') + '/' + TRAINING_DATA)
df2 = df.sample(50000)
X = df2['question_text']
Y = df2['target']
Xtrain, Xtest, Ytrain, Ytest = train_test_split(X,
Y,
test_size=0.2,
random_state=42)
# Two classes - Fake=0, Reliable=1
config.set('num_target_tokens', 2)
classifier = ClassifierFactory.getSVM()
print('training size: ', len(Xtrain))
print('testing size: ', len(Xtest))
print('start fitting ...')
classifier.fit(Xtrain, Ytrain, Xtest, Ytest)
df = pd.read_csv(Config.getPath('data') + '/' + TESTING_DATA)
df = df.sample(100000)
X = df['question_text']
Y = df['target']
pred = classifier.predict(X)
score = metrics.accuracy_score(Y, pred)
f1score = metrics.f1_score(Y, pred)
print("accuracy: %0.3f" % score)
print("f1 score: %0.3f" % f1score)
cm = metrics.confusion_matrix(Ytest, pred, labels=[0, 1])
plot_confusion_matrix(cm, classes=[0, 1])
def load_config(**kwargs):
print('loading csv file ...')
global config
df = pd.read_csv(Config.getPath('data') + '/' + TRAINING_DATA)
df = df.sample(50000)
X = df['question_text']
Y = df['target']
print('preparing configuration...')
config = ClassifierFactory.getConfig(X, Y, json_file=None, **kwargs)
#Two classes - Fake=0, Reliable=1
config.set('num_target_tokens', 2)
def train_vanilla(classifier):
global config
print('configuration extracted from input texts ...')
Xtrain, Xtest, Ytrain, Ytest = train_test_split(config.getData(),
config.getLabels(),
test_size=0.2,
random_state=42)
print('training size: ', len(Xtrain))
print('testing size: ', len(Xtest))
print('start fitting ...')
history = classifier.fit(Xtrain,
Ytrain,
Xtest,
Ytest,
epochs=config.get('epochs'))
if (history != None):
history_plot_file_path = Config.getPath(
'reports') + '/' + classifier.model_name + '-history.png'
plot_and_save_history(history, classifier.model_name,
history_plot_file_path)
def train_experiment(classifier):
print('loading csv file ...')
global config
df = pd.read_csv(Config.getPath('data') + '/' + TRAINING_DATA)
df = df.sample(20000)
X = df['question_text']
Y = df['target']
print('splitting data...')
Xtrain, Xtest, Ytrain, Ytest = train_test_split(X,
Y,
test_size=0.2,
random_state=42)
print('training size: ', len(Xtrain))
print('testing size: ', len(Xtest))
print('start fitting ...')
# max_sequence, vocab_size, lstm_units, dropout
experiment = [
[20, 5000, 64, 0.2], #0
[35, 5000, 64, 0.2], #1
[50, 5000, 64, 0.2], #2
[100, 5000, 64, 0.2], #3
[50, 5000, 128, 0.2], # 4
[50, 5000, 256, 0.2], # 5
[50, 5000, 512, 0.2], # 6
[50, 2000, 64, 0.2], # 7
[50, 3000, 64, 0.2], # 8
[50, 4000, 64, 0.2], # 9
[50, 5000, 64, 0.2], # 10
[50, 6000, 64, 0.2], # 11
[50, 7000, 64, 0.2], # 12
[50, 8000, 64, 0.2], # 13
[50, 9000, 64, 0.2], # 14
[50, 5000, 64, 0.1], # 15
[50, 5000, 64, 0.2], # 16
[50, 5000, 64, 0.3], # 17
[50, 5000, 64, 0.4], # 18
]
i = 0
for max_seq, vocab_siz, lstm_u, drop in experiment:
config = Config(X, Y, max_seq, vocab_siz, 2, 'glove')
print("%s starting experiment ... %d" % (datetime.datetime.now(), i))
#model=ClassifierFactory.getLSTM(**{'config':config,'lstm_units':lstm_u,'dropout':drop})
model = ClassifierFactory.getLSTM(**{
'config': config,
'lstm_units': lstm_u,
'dropout': drop
})
history = model.fit(Xtrain,
Ytrain,
Xtest,
Ytest,
epochs=10,
file_prefix='experiment-%i' % i)
history_plot_file_path = Config.getPath(
'reports') + '/' + model.model_name + ('_experiment_%d' %
i) + '-history.png'
plot_and_save_history(history, model.model_name,
history_plot_file_path)
i += 1