test_indices, [],
price_array_training,
price_array, **cnn_kwargs)
print(fitting_dictionary['fitted_strategy_score'])
if len(portfolio_allocations) != 0:
portfolio_allocations = np.vstack([
portfolio_allocations,
fitting_dictionary['fitted_strategy_score']
])
else:
portfolio_allocations = fitting_dictionary['fitted_strategy_score']
plot_portfolio_value, cum_log_return = calculate_portfolio_value(
portfolio_allocations, price_array[initial_idx:(idx + 1), :],
liquidation_factor[(initial_idx + 1):(idx + 1), :])
print('Backtest step ', idx - initial_idx, ' of ', run_length)
if len(plot_portfolio_value) != 0:
print('Cumulative profit fraction:', plot_portfolio_value[-1] - 1)
print()
if plot_time:
plot_time.append(plot_time[-1] + 1 / 48.0)
else:
plot_time.append(1 / 48.0)
plot_prices = np.cumprod(price_array[initial_idx:(idx + 1), 0], axis=0)
other_prices = np.cumprod(price_array[initial_idx:(idx + 1), :],
# [0.9, 0.05, 0.0001])
train_indices = np.linspace(0, len(price_array_training)-100, len(price_array_training)-100, dtype=int)
test_indices = np.linspace(len(price_array_training) - 100, len(price_array_training) - 75, 25, dtype=int)
validation_indices = np.linspace(len(price_array_training) -5, len(price_array_training)-1, 5, dtype=int)
fitting_dictionary = tensorflow_cnn_fitting(
train_indices,
test_indices,
validation_indices,
price_array_training,
price_array,
load_net='model.h5')
portfolio_value, cum_log_return = calculate_portfolio_value(
fitting_dictionary['fitted_strategy_score'],
price_array[test_indices, :])
final_portfolio_value = portfolio_value[-1]
running_portfolio_value *= final_portfolio_value
print('Run profit fraction: ', final_portfolio_value - 1)
print('Offset tests cumulative profit fraction:', running_portfolio_value - 1)
from matplotlib import pyplot as plt
plt.figure()
plt.plot(portfolio_value)
plt.xlabel('Time (30 minute steps)')
plt.ylabel('Fractional Portfolio Value')
plt.figure()