def predict_test(vectorizer, model):
fl = 'C:\\Users\\prach\\Desktop\\Political-Opinion-Mining\\Test Data Pre Processing\\test_labelled.csv'
data = pd.read_csv(fl, encoding='utf-8-sig')
data = data[['sentiment', 'text']].dropna()
tweets = data['text']
y = data['sentiment'].to_numpy()
tweets = [preprocess(tweet) for tweet in tweets]
X_test_vec = vectorizer.transform(tweets)
yhat = predict(model, X_test_vec)
yhat = np.array(yhat)
print(y)
print(yhat)
print((yhat == y).mean())
fpr, tpr, thresholds = roc(y, model.predict_proba(X_test_vec)[:, 0])
print('fpr', fpr)
print('tpr', tpr)
print('thresholds', thresholds)
# plot the roc curve for the model
pyplot.plot(fpr, tpr, marker='.', label='Gradient Boost')
# axis labels
pyplot.xlabel('False Positive Rate')
pyplot.ylabel('True Positive Rate')
# show the legend
pyplot.legend()
# show the plot
# pyplot.show()
pyplot.savefig('gdbt_test.png')
def predict_test(vectorizer, model):
fl = 'C:\\Users\\prach\\Desktop\\Political-Opinion-Mining\\Test Data Pre Processing\\test_labelled.csv'
data = pd.read_csv(fl, encoding='utf-8-sig')
data = data[['sentiment', 'text']].dropna()
tweets = data['text']
y = data['sentiment'].to_numpy()
tweets = [preprocess(tweet) for tweet in tweets]
yhat = predict(model, vectorizer.transform(tweets))
yhat = np.array(yhat)
print(y)
print(yhat)
print((yhat == y).mean())
from matplotlib import pyplot
from supervised.ml import preprocess_data, load_train_data_2, unpickle_model, train_lr, evaluate, pickle_model, \
predict, confusion_matrix, roc
if __name__ == '__main__':
X_train, y_train, X_test, y_test = preprocess_data(load_train_data_2())
vectorizer = unpickle_model('tfidf.sav')
X_train_vec = vectorizer.transform(X_train)
X_test_vec = vectorizer.transform(X_test)
lr = train_lr(X_train_vec, y_train, C=10)
acc_train = evaluate(predict(lr, X_train_vec), y_train)
acc_test = evaluate(predict(lr, X_test_vec), y_test)
conf_mat = confusion_matrix(predict(lr, X_test_vec), y_test)
# fpr, tpr, thresholds = roc(y_test, lr.predict_proba(X_test_vec)[:, 0])
# print('fpr', fpr)
# print('tpr', tpr)
# print('thresholds', thresholds)
#
# # plot the roc curve for the model
#
# pyplot.plot(fpr, tpr, marker='.', label='Logistic')
# # axis labels
# pyplot.xlabel('False Positive Rate')
# pyplot.ylabel('True Positive Rate')
# # show the legend
# pyplot.legend()
# # show the plot
# pyplot.show()
# print(f'Train Acc = {acc_train} Test Acc = {acc_test}')
from supervised.ml import preprocess_data, load_train_data, unpickle_model, evaluate, pickle_model, \
train_gbdt, predict
if __name__ == '__main__':
X_train, y_train, X_test, y_test = preprocess_data(load_train_data())
vectorizer = unpickle_model('tfidf.sav')
X_train_vec = vectorizer.transform(X_train)
X_test_vec = vectorizer.transform(X_test)
model = train_gbdt(X_train_vec, y_train, n_est=20000)
acc_train = evaluate(predict(model, X_train_vec), y_train)
acc_test = evaluate(predict(model, X_test_vec), y_test)
print(f'Train Acc = {acc_train} Test Acc = {acc_test}')
# n_est = 100: Train Acc = 0.5661953125 Test Acc = 0.56625625
# n_est = 1000: Train Acc = 0.71187109375 Test Acc = 0.708015625
pickle_model(model, 'gbdt20000.sav')
model = train_gbdt(X_train_vec, y_train, n_est=30000)
acc_train = evaluate(predict(model, X_train_vec), y_train)
acc_test = evaluate(predict(model, X_test_vec), y_test)
print(f'Train Acc = {acc_train} Test Acc = {acc_test}')
# n_est = 100: Train Acc = 0.5661953125 Test Acc = 0.56625625
# n_est = 1000: Train Acc = 0.71187109375 Test Acc = 0.708015625
# n_est = 5000: Train Acc = 0.78938203125 Test Acc = 0.7805875
# n_est = 10000 Train Acc = 0.8103390625 Test Acc = 0.797578125
# n_est = 15000 Train Acc = 0.82
# n_est = 20000 Train Acc = 0.829825 Test Acc = 0.810390625
pickle_model(model, 'gbdt30000.sav')
from supervised.ml import preprocess_data, load_train_data, unpickle_model, train_lr, evaluate, pickle_model, \
predict
if __name__ == '__main__':
X_train, y_train, X_test, y_test = preprocess_data(load_train_data())
vectorizer = unpickle_model('tfidf.sav')
X_train_vec = vectorizer.transform(X_train)
X_test_vec = vectorizer.transform(X_test)
lr = train_lr(X_train_vec, y_train)
acc_train = evaluate(predict(lr, X_train_vec), y_train)
acc_test = evaluate(predict(lr, X_test_vec), y_test)
print(f'Train Acc = {acc_train} Test Acc = {acc_test}')
# Train Acc = 0.853759375 Test Acc = 0.824384375
pickle_model(lr, 'lr.sav')