def __init__(self, nr_inp, nr_out):
self.n_nodes_input = nr_inp
self.n_nodes_output = nr_out
self.x = tf.placeholder('float', [None, nr_inp])
self.y = tf.placeholder('float', [None, nr_out])
reading_from_file.read_from_file_time_series_norwegian(
self.n_nodes_input, self.n_nodes_output)
reading_from_file.normalize_time_series()
self.batch_size = reading_from_file.get_test_data_size_time_series()
self.batch_size = 1000
def test_persistnece_time():
reading_from_file.read_from_file_time_series_norwegian(5, 5)
reading_from_file.normalize_time_series()
test_output = reading_from_file.get_test_output_time_series()
test_input = reading_from_file.get_test_input_time_series()
result = persistence_test.persistence_test(test_input, test_output)
print("Result:" + str(result))
with open('output_files/raggovidda_persistence_test_time_norwegian.csv',
'w') as csvfile:
spamwriter = csv.writer(csvfile,
delimiter=';',
quotechar='|',
quoting=csv.QUOTE_MINIMAL)
spamwriter.writerow(result)
def run_svr():
#reading_from_file.read_from_file_time_series(5, 5)
reading_from_file.read_from_file_time_series_norwegian(5, 5)
reading_from_file.normalize_time_series()
test_input = reading_from_file.get_test_input_time_series()
test_output = reading_from_file.get_test_output_time_series()
train_input = reading_from_file.get_train_input_time_series()
train_output = reading_from_file.get_train_output_time_series()
result = svr.predict_stuff(train_input, train_output, test_input,
test_output)
with open('output_files/raggovidda_svr_norwegian.csv', 'w') as csvfile:
spamwriter = csv.writer(csvfile,
delimiter=';',
quotechar='|',
quoting=csv.QUOTE_MINIMAL)
spamwriter.writerow(result)
def __init__(self, nr_inp, nr_out):
self.n_nodes_input = nr_inp
self.n_noes_actial_out = nr_out
self.n_nodes_output = 1
self.x = tf.placeholder('float', [None, nr_inp])
self.y = tf.placeholder('float', [None, self.n_nodes_output])
reading_from_file.read_from_file_time_series_norwegian(
self.n_nodes_input, self.n_noes_actial_out)
reading_from_file.normalize_time_series()
#self.batch_size = reading_from_file.get_test_data_size_time_series()
self.batch_size = 1000
self.train_outputs = reading_from_file.get_train_output_time_series()
lenght = (len(self.train_outputs))
self.columns = list(zip(*self.train_outputs))
self.output_values = self.columns[0]
self.output_values = np.asarray(self.output_values).reshape(lenght, 1)
def test_knn_time_new():
#reading_from_file.read_from_file_time_series(5, 5)
reading_from_file.read_from_file_time_series_norwegian(5, 5)
reading_from_file.normalize_time_series()
test_input = reading_from_file.get_test_input_time_series()
test_output = reading_from_file.get_test_output_time_series()
train_input = reading_from_file.get_train_input_time_series()
train_output = reading_from_file.get_train_output_time_series()
with open('output_files/raggovidda_knn_norwegian.csv', 'w') as csvfile:
spamwriter = csv.writer(csvfile,
delimiter=';',
quotechar='|',
quoting=csv.QUOTE_MINIMAL)
for k in range(1, 30):
result = knn_improved.knn_new(train_input, train_output,
test_input, test_output, k)
print("Result:" + str(result) + " For K = " + str(k))
spamwriter.writerow(result)
def lstm_keras(nr_input, nr_output):
reading_from_file.read_from_file_time_series_norwegian(nr_input, nr_output)
reading_from_file.normalize_time_series()
train_inp = reading_from_file.get_train_input_time_series()
train_out = reading_from_file.get_train_output_time_series()
x_train, y_train, x_test, y_test = lstm.load_data(
'test_files/Raggovidda_2.csv', 5, True)
#print("X:", x_test)
#print("Y:", y_test)
print(len(y_test))
print(len(x_test))
train_inp = np.asarray(train_inp)
len_inp = len(train_inp)
train_inp = train_inp.reshape(len_inp, 5, 1)
columns = list(zip(*train_out))
train_out = columns[0]
train_out = np.reshape(train_out, len(train_out))
test_inp = reading_from_file.get_test_input_time_series()
test_out = reading_from_file.get_test_output_time_series()
columns = list(zip(*test_out))
test_out = columns[0]
test_out = np.reshape(test_out, len(test_out))
len_inp = len(test_inp)
test_inp = test_inp.reshape(len_inp, 5, 1)
#test_out = np.asarray(test_out)
#test_out = test_out.reshape(len_inp, 5, 1)
model = Sequential()
model.add(LSTM(input_dim=1, output_dim=5, return_sequences=True))
model.add(Dropout(0.2))
model.add(LSTM(100, return_sequences=False))
model.add(Dropout(0.2))
model.add(Dense(output_dim=1))
model.add(Activation('linear'))
start = time.time()
model.compile(loss='mse', optimizer='rmsprop')
print('compilation time: ', time.time() - start)
model.fit(x_train,
y_train,
batch_size=2000,
validation_split=0.05,
epochs=5)
predictions = lstm.predict_sequences_multiple(model, x_test, 5, 5)
pred = np.asarray(predictions)
y_val = np.asarray(y_test)
y_val = y_val.reshape(len(pred), 5)
#print ("Y_val: ",y_val)
#print ("Pred: ", pred)
print(len(y_val))
print(len(pred))
result = mean_squared_error(pred, y_val, multioutput='raw_values')
print(result)
return result