def model3():
# 3 модель прогнозирования "В лоб"(слой с обратным распространением)
from defs_neuro import fit_lstm_bidir
n_lag = 1
n_seq = 1
n_test = 7
n_epochs = 200
n_batch = 1
n_neurons = 3
scaler, train, test = prepare_data(series, n_test, n_lag, n_seq)
model = fit_lstm_bidir(train, n_lag, n_seq, n_batch, n_epochs,
n_neurons)
model.reset_states()
forecasts = make_forecasts(model, n_batch, test, n_lag, n_seq)
forecasts = inverse_transform(series, forecasts, scaler, n_test - 1)
actual = [row[n_lag:] for row in test]
actual = inverse_transform(series, actual, scaler, n_test - 1)
evaluate_forecasts(actual, forecasts, n_lag, n_seq)
plot_forecasts(series[(row_count - 12):(row_count)], forecasts,
n_test - 1)
def model5():
from numpy import vstack
from defs_neuro import prepare_data
n_lag = 1
n_seq = 1
n_test = 7
n_epochs = 200
n_batch = 1
n_neurons = 4
scaler, train, test = prepare_data(series, n_test, n_lag, n_seq)
scaler_1, train_1, test_1 = prepare_data(series, n_test, n_lag, n_seq)
scaler_2, train_2, test_2 = prepare_data(values, n_test, n_lag, n_seq)
train = vstack((train_1[:, 0], train_2[:, 0], train_1[:, 1]))
train = train.T
model = fit_lstm(train, n_lag + 1, n_seq, n_batch, n_epochs, n_neurons)
model.reset_states()
test = vstack((test_1[:, 0], test_2[:, 0], test_1[:, 1]))
test = test.T
forecasts = make_forecasts(model, n_batch, test, n_lag + 1, n_seq)
forecasts = inverse_transform(series, forecasts, scaler, n_test)
actual = [row[n_lag + 1:] for row in test]
actual = inverse_transform(series, actual, scaler, n_test)
evaluate_forecasts(actual, forecasts, n_lag, n_seq)
plot_forecasts(series[(row_count - 12):(row_count)], forecasts, n_test)
def model2():
from defs_neuro import fit_lstm_stack
n_lag = 1
n_seq = 1
n_test = 7
n_epochs = 200
n_batch = 1
n_neurons = 4
scaler, train, test = prepare_data(series, n_test, n_lag, n_seq)
model = fit_lstm_stack(train, n_lag, n_seq, n_batch, n_epochs,
n_neurons)
model.reset_states()
forecasts = make_forecasts(model, n_batch, test, n_lag, n_seq)
forecasts = inverse_transform(series, forecasts, scaler, n_test)
actual = [row[n_lag:] for row in test]
actual = inverse_transform(series, actual, scaler, n_test)
evaluate_forecasts(actual, forecasts, n_lag, n_seq)
plot_forecasts(series[(row_count - 12):(row_count)], forecasts, n_test)
def model1():
n_lag = 1
n_seq = 1
n_test = 7
n_epochs = 200
n_batch = 1
n_neurons = 4
scaler, train, test = prepare_data(series, n_test, n_lag, n_seq)
model = fit_lstm(train, n_lag, n_seq, n_batch, n_epochs, n_neurons)
model.reset_states()
forecasts = make_forecasts(model, n_batch, test, n_lag, n_seq)
forecasts = inverse_transform(series, forecasts, scaler, n_test)
actual = [row[n_lag:] for row in test]
actual = inverse_transform(series, actual, scaler, n_test)
evaluate_forecasts(actual, forecasts, n_lag, n_seq)
name = 'Модель 1'
plot_forecasts(series[rc2:rc], forecasts, n_test, name)