def prediction():
filepath = request.files['csv_file'].read()
model = request.files['model'].read()
df = pd.read_csv(filepath)
dt = list()
for i in range(len(df)):
dt.append(df['Date'][i] + " " + df['Time'][i])
df['DateTime'] = dt
df['DateTime'] = pd.to_datetime(df['DateTime'])
df.index = df['DateTime']
df.drop(df[df['Volume'] == 0].index, axis=0, inplace=True)
idx = df[df['Low'] == df['High']].index
df.drop(idx, axis=0, inplace=True)
df['Date'] = pd.to_datetime(df['Date'])
df.index = df['Date']
data = df['Close'].copy()
# Training and testing data
train_size = int(len(data) * 0.80)
train = data[:train_size]
# test = data[train_size :]
# Live Testing
test_size = 60
test = data[len(df) - test_size:]
# Data scaling
scaler = MinMaxScaler()
scaled_train = scaler.fit_transform(np.array(train).reshape(-1, 1))
# Test
scaled_test = scaler.transform(np.array(test).reshape(-1, 1))
X_test = [i for i in test]
X_test = np.array(X_test)
regressor = SupervisedDBNRegression.load(model)
y_pred = regressor.predict(X_test)
# print('Metrices of Regressor ')
# print('\nR-squared: %f\nMSE: %f' % (r2_score(y_test, y_pred), mean_squared_error(y_test, y_pred)))
# print('Accuracy = {}'.format(r2_score(y_test , y_pred) * 100))
# y_test = scaler.inverse_transform(y_test.reshape(-1,1))
y_pred = scaler.inverse_transform(y_pred.reshape(-1, 1))
# plt.figure(figsize=(20,9))
# # plt.plot(train.index , train , label = 'Training data')
# plt.plot(test.index[60:] , y_test , label='Test' , color ='b' )
# plt.plot(test.index[60:], y_pred , label='Regressor Predictions' , color='g')
# plt.title('Actual v/s Predicted')
# plt.legend()
# plt.savefig('./assets/output.png')
# print(y_pred)
return jsonify({'Prediction': y_pred[0][0]})
def build_RBM(num_bp, epoch_pretrain=125, batch_size=24, hidDim=[100, 140]):
regressor = SupervisedDBNRegression(hidden_layers_structure=hidDim,
learning_rate_rbm=0.01,
learning_rate=0.01,
n_epochs_rbm=epoch_pretrain,
n_iter_backprop=num_bp,
batch_size=batch_size,
activation_function='sigmoid')
return regressor
def testing(csv_filepath, save_output_image, model_path):
df = pd.read_csv(csv_filepath)
dt = list()
for i in range(len(df)):
dt.append(df['Date'][i] + " " + df['Time'][i])
df['DateTime'] = dt
df['DateTime'] = pd.to_datetime(df['DateTime'])
df.index = df['DateTime']
df.drop(df[df['Volume'] == 0].index, axis=0, inplace=True)
idx = df[df['Low'] == df['High']].index
df.drop(idx, axis=0, inplace=True)
df['Date'] = pd.to_datetime(df['Date'])
df.index = df['Date']
data = df['Close'].copy()
# Training and testing data
train_size = int(len(data) * 0.80)
train = data[:train_size]
# Live Testing
test_size = 60
test = data[len(df) - test_size:]
# Data scaling
scaler = MinMaxScaler()
scaled_train = scaler.fit_transform(np.array(train).reshape(-1, 1))
# Test
scaled_test = scaler.transform(np.array(test).reshape(-1, 1))
X_test = [i for i in test]
X_test = np.array(X_test)
try:
FILE_PATH = model_path
regressor = SupervisedDBNRegression.load(FILE_PATH)
except:
print('Unable to load model')
y_pred = regressor.predict(X_test)
print('Predicted value is {}'.format(y_pred[0][0]))
'/grafico_' + os.path.splitext(conjTreino)[0])
plt.close()
path1 = './Modelos/'
# path2 = 'Treinamento_2019_10_11_14_40_41/'
# path3 = 'modelo_treinamento_2019_10_11_14_40_41_modelo_101_[90, 90, 90].pkl'
path2 = 'Treinamento_2019_11_01_14_10_48/'
path3 = 'modelo_treinamento_2019_11_01_14_10_48_modelo_1_[90, 90, 90].pkl'
pathCompleto = path1 + path2 + path3
regressor = SupervisedDBNRegression.load(pathCompleto)
# Teste
Y_pred = regressor.predict(X_test)
# if conjTreino == 'degrauUnitario.csv':
# Y_pred = Y_pred / 4.6 # 4.62073146825719
r2Score = r2_score(Y_test, Y_pred)
MSE = mean_squared_error(Y_test, Y_pred)
print('\nDone.\nR-squared: %f\nMSE: %f' % (r2Score, MSE))
arquivoResultados = pd.DataFrame(data={
"Arquivo": [conjTreino],
# scaler = MinMaxScaler()
# scaled_train = scaler.fit_transform(np.array(train).reshape(-1,1))
# log_change_train = np.log(train).pct_change().dropna()
# log_change_test = np.log(test).pct_change().dropna()
# print(log_change_train)
# Preparing training data
scaled_train = np.array(train)
X_train, y_train = [], []
for i in range(60, len(scaled_train)):
X_train.append(scaled_train[i - 60:i])
y_train.append(scaled_train[i])
# Training
X_train = np.array(X_train)
y_train = np.array(y_train)
regressor = SupervisedDBNRegression(hidden_layers_structure=[100],
learning_rate_rbm=0.001,
learning_rate=0.001,
n_epochs_rbm=60,
n_iter_backprop=100,
batch_size=32,
activation_function='relu')
regressor.fit(X_train, y_train)
print('*' * 15, ' Model is saving ', '*' * 15)
regressor.save('./models/nifty_24_03_2021.pkl')
print('*' * 15, ' Model saved ', '*' * 15)
def testing(csv_filepath , save_output_image , model_path):
df = pd.read_csv(csv_filepath)
dt = list()
for i in range(len(df)):
dt.append(df['Date'][i]+ " " +df['Time'][i])
df['DateTime'] = dt
df['DateTime'] = pd.to_datetime(df['DateTime'])
df.index = df['DateTime']
df.drop(df[df['Volume']==0].index , axis=0 ,inplace=True)
idx = df[df['Low']==df['High']].index
df.drop(idx , axis=0 , inplace=True)
df['Date'] = pd.to_datetime(df['Date'])
df.index = df['Date']
data = df['Close'].copy()
# Training and testing data
train_size = int(len(data) * 0.80)
train = data[:train_size]
test = data[train_size :]
# Data scaling
scaler = MinMaxScaler()
scaled_train = scaler.fit_transform(np.array(train).reshape(-1,1))
# Test
scaled_test = scaler.transform(np.array(test).reshape(-1,1))
X_test , y_test = [] , []
for i in range(60 , len(scaled_test)):
X_test.append(scaled_test[i-60: i,0])
y_test.append(scaled_test[i ,0])
X_test = np.array(X_test)
y_test = np.array(y_test)
print(X_test.shape)
try:
FILE_PATH = model_path
regressor = SupervisedDBNRegression.load(FILE_PATH)
print('Model loaded Successfully ')
except:
print('Unable to load model')
try:
y_pred = regressor.predict(X_test)
print('Metrices of Regressor ')
print('\nR-squared: %f\nMSE: %f' % (r2_score(y_test, y_pred), mean_squared_error(y_test, y_pred)))
print('Accuracy = {}'.format(r2_score(y_test , y_pred) * 100))
try:
y_test = scaler.inverse_transform(y_test.reshape(-1,1))
y_pred = scaler.inverse_transform(y_pred.reshape(-1,1))
plt.figure(figsize=(20,9))
# plt.plot(train.index , train , label = 'Training data')
plt.plot(test.index[60:] , y_test , label='Test' , color ='b' )
plt.plot(test.index[60:], y_pred , label='Regressor Predictions' , color='g')
plt.title('Actual v/s Predicted')
plt.legend()
if save_output_image==True:
plt.savefig('./assets/output.png')
plt.show()
except Exception as e:
print(e)
except Exception as e:
print(e)
# print(log_change_train)
# Test
# scaled_test = scaler.transform(np.array(test).reshape(-1,1))
X_test, y_test = [], []
for i in range(60, len(scaled_test)):
X_test.append(scaled_test[i - 60:i, 0])
y_test.append(scaled_test[i, 0])
X_test = np.array(X_test)
y_test = np.array(y_test)
# print(X_test.shape)
try:
FILE_PATH = '/home/rahul/Desktop/dbn/nifty_20_03_2021.pkl'
regressor = SupervisedDBNRegression.load(FILE_PATH)
print('Model loaded Successfully ')
except:
print('Unable to load model')
try:
y_pred = regressor.predict(X_test)
print('\nR-squared: %f\nMSE: %f' %
(r2_score(y_test, y_pred), mean_squared_error(y_test, y_pred)))
try:
y_test = scaler.inverse_transform(y_test.reshape(-1, 1))
y_pred = scaler.inverse_transform(y_pred.reshape(-1, 1))
# y_test = return_log_normal(y_test)
# y_pred = return_log_normal(y_pred)
plt.figure(figsize=(20, 9))
X.append(cek)
y.append(data[i + lag])
return np.array(X), np.array(y)
lag = 3
X_train, y_train = create_sliding_windows(train, lag)
y_train = np.reshape(y_train, (len(y_train), 1))
X_val, y_val = create_sliding_windows(validation, lag)
X_test, y_test = create_sliding_windows(test, lag)
# Training
regressor = SupervisedDBNRegression(hidden_layers_structure=[30, 40],
learning_rate_rbm=0.001,
learning_rate=0.001,
n_epochs_rbm=30,
n_iter_backprop=100,
batch_size=32,
activation_function='relu',
dropout_p=0.2)
regressor.fit(X_train, y_train)
# Save the model
regressor.save('model.pkl')
# Restore it
regressor = SupervisedDBNRegression.load('model.pkl')
# Test
y_pred = regressor.predict(X_test)
print('Done.\nMAE: %f' % mean_absolute_error(y_test, y_pred))
Y_train = np.array(train_data[0:316, [30]])
# 提取测试集特征列
# test_x = np.array(test_data[0:7, [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29]])
X_test = np.array(test_data[0:1000, [
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20,
21, 22, 23, 24, 25, 26, 27, 28, 29
]])
#Y_test = np.array(train_data[0:6, [30]])
# Data scaling
min_max_scaler = MinMaxScaler()
X_train = min_max_scaler.fit_transform(X_train)
# Training
regressor = SupervisedDBNRegression(hidden_layers_structure=[130],
learning_rate_rbm=0.05,
learning_rate=0.05,
n_epochs_rbm=20,
n_iter_backprop=20,
batch_size=16,
activation_function='sigmoid')
regressor.fit(X_train, Y_train)
# Test
X_test = min_max_scaler.transform(X_test)
Y_pred = regressor.predict(X_test)
print("测试集预测结果")
print(Y_pred)
#print('Done.\nR-squared: %f\nMSE: %f' % (r2_score(Y_test, Y_pred), mean_squared_error(Y_test, Y_pred)))
# Loading dataset
boston = load_boston()
X, Y = boston.data, boston.target
# Splitting data
X_train, X_test, Y_train, Y_test = train_test_split(X,
Y,
test_size=0.2,
random_state=0)
# Data scaling
min_max_scaler = MinMaxScaler()
X_train = min_max_scaler.fit_transform(X_train)
# Training
regressor = SupervisedDBNRegression(hidden_layers_structure=[200],
learning_rate_rbm=0.01,
learning_rate=0.01,
n_epochs_rbm=100,
n_iter_backprop=500,
l2_regularization=0.0,
batch_size=32,
activation_function='relu')
regressor.fit(X_train, Y_train)
# Test
X_test = min_max_scaler.transform(X_test)
Y_pred = regressor.predict(X_test)
print 'Done.\nR-squared: %f\nMSE: %f' % (r2_score(
Y_test, Y_pred), mean_squared_error(Y_test, Y_pred))