def random_search(strategy_dictionary_local, n_iterations):
toc = tic()
counter = 0
error = 1e10
while counter < n_iterations:
counter += 1
#strategy_dictionary['sequence_flag'] = np.random.choice([True, False])
if strategy_dictionary['sequence_flag']:
strategy_dictionary_local = randomise_sequence_dictionary_inputs(strategy_dictionary_local)
else:
strategy_dictionary_local = randomise_dictionary_inputs(strategy_dictionary_local)
fitting_dictionary, data_to_predict, error_loop, profit_factor = fit_tensorflow(strategy_dictionary_local)
if error_loop < error:
error = error_loop
strategy_dictionary_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_optimum, fitting_dictionary_optimum, data_to_predict, toc)
return strategy_dictionary_optimum
def random_search(strategy_dictionary_local, n_iterations):
toc = tic()
data_local, data_2 = import_data(strategy_dictionary)
fitting_inputs_local, continuous_targets, classification_targets = input_processing(
data_local, data_2, strategy_dictionary)
counter = 0
error = 1e5
while counter < n_iterations:
counter += 1
strategy_dictionary_local = randomise_dictionary_inputs(strategy_dictionary_local)
if strategy_dictionary['regression_mode'] == 'classification':
fitting_targets_local = classification_targets
elif strategy_dictionary['regression_mode'] == 'regression':
fitting_targets_local = continuous_targets
fitting_inputs_local = preprocessing_inputs(strategy_dictionary, fitting_inputs_local)
fitting_dictionary, profit_factor = fit_strategy(
strategy_dictionary, data_local, fitting_inputs_local, fitting_targets_local)
error_loop = fitting_dictionary['error']
if error_loop < error and fitting_dictionary['n_trades'] != 0:
error = error_loop
strategy_dictionary_local_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_local_optimum, fitting_dictionary_optimum, data_local, toc)
return strategy_dictionary_local_optimum, fitting_inputs_local, fitting_targets_local, data_local
def random_search(strategy_dictionary_local, n_iterations):
print("Ticking")
toc = tic()
print("Importing data")
data_local, data_2 = import_data(strategy_dictionary_local)
print("Finished Imported data")
fitting_inputs_local, continuous_targets, classification_targets = input_processing(
data_local, data_2, strategy_dictionary)
print("Targets determined")
counter = 0
error = 1e5
fitting_dictionary_optimum = []
strategy_dictionary_optimum = []
fitting_targets_local = []
while counter < n_iterations:
print("Starting iteration %s" % counter)
counter += 1
# strategy_dictionary['sequence_flag'] = np.random.choice([True, False])
strategy_dictionary['sequence_flag'] = False
if strategy_dictionary['sequence_flag']:
strategy_dictionary_local = randomise_sequence_dictionary_inputs(
strategy_dictionary_local)
else:
strategy_dictionary_local = randomise_dictionary_inputs(
strategy_dictionary_local)
if strategy_dictionary_local['regression_mode'] == 'classification':
fitting_targets_local = classification_targets
elif strategy_dictionary_local['regression_mode'] == 'regression':
fitting_targets_local = continuous_targets
fitting_inputs_local, strategy_dictionary_local = preprocessing_inputs(
strategy_dictionary_local, fitting_inputs_local)
fitting_dictionary, error_loop, profit_factor = fit_tensorflow(
strategy_dictionary_local, data_local, fitting_inputs_local,
fitting_targets_local)
if error_loop < error:
error = error_loop
strategy_dictionary_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
print("Completed iteration")
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_optimum,
fitting_dictionary_optimum, data_local, toc)
return strategy_dictionary_optimum, data_local, fitting_inputs_local, fitting_targets_local
def fit_time_scale(strategy_dictionary_input, search_iterations_local,
time_iterations):
""" fit timescale variables"""
toc = tic()
counter = 0
strategy_dictionary_optimum = []
optimum_profit = -2
while counter < time_iterations:
strategy_dictionary_input = randomise_time_inputs(
strategy_dictionary_input)
strategy_dictionary_local,\
fitting_dictionary_local,\
fitting_inputs_local,\
fitting_targets_local,\
data_local,\
test_profit\
= random_search(
strategy_dictionary_input,
search_iterations_local,
toc)
if test_profit > optimum_profit:
optimum_profit = test_profit
strategy_dictionary_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary_local
counter += 1
underlined_output('Best strategy fit')
if strategy_dictionary['plot_last']:
strategy_dictionary['plot_flag'] = True
output_strategy_results(strategy_dictionary_optimum,
fitting_dictionary_optimum,
data_local,
toc,
momentum_dict=simple_momentum_comparison(
data_local, strategy_dictionary_optimum,
fitting_dictionary_optimum))
return strategy_dictionary_optimum,\
fitting_inputs_local,\
fitting_targets_local,\
data_local
def random_search(strategy_dictionary_local, n_iterations):
toc = tic()
data_local, data_2 = import_data(strategy_dictionary_local)
fitting_inputs_local, continuous_targets, classification_targets = input_processing(
data_local, data_2, strategy_dictionary)
counter = 0
error = 1e5
fitting_dictionary_optimum = []
strategy_dictionary_optimum = []
fitting_targets_local = []
while counter < n_iterations:
counter += 1
strategy_dictionary['sequence_flag'] = np.random.choice([True, False])
if strategy_dictionary['sequence_flag']:
strategy_dictionary_local = randomise_sequence_dictionary_inputs(
strategy_dictionary_local)
else:
strategy_dictionary_local = randomise_dictionary_inputs(
strategy_dictionary_local)
if strategy_dictionary_local['regression_mode'] == 'classification':
fitting_targets_local = classification_targets
elif strategy_dictionary_local['regression_mode'] == 'regression':
fitting_targets_local = continuous_targets
fitting_inputs_local, strategy_dictionary_local = preprocessing_inputs(
strategy_dictionary_local, fitting_inputs_local)
fitting_dictionary, error_loop, profit_factor = fit_tensorflow(
strategy_dictionary_local, data_local, fitting_inputs_local,
fitting_targets_local)
if error_loop < error:
error = error_loop
strategy_dictionary_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_optimum,
fitting_dictionary_optimum, data_local, toc)
return strategy_dictionary_optimum, data_local, fitting_inputs_local, fitting_targets_local
def random_search(strategy_dictionary_local, n_iterations):
"""random search to find optimum machien learning algorithm and preprocessing"""
toc = tic()
data_local = import_data(strategy_dictionary_local)
fitting_inputs_local, continuous_targets, classification_targets = input_processing(
data_local, strategy_dictionary_local)
counter = 0
error = 1e5
fitting_targets_local = []
fitting_dictionary_optimum = []
strategy_dictionary_optimum = []
while counter < n_iterations:
counter += 1
strategy_dictionary_local = randomise_dictionary_inputs(
strategy_dictionary_local)
if strategy_dictionary_local['regression_mode'] == 'classification':
fitting_targets_local = classification_targets.astype(int)
elif strategy_dictionary_local['regression_mode'] == 'regression':
fitting_targets_local = continuous_targets
fitting_inputs_local, strategy_dictionary = preprocessing_inputs(
strategy_dictionary_local, fitting_inputs_local)
fitting_dictionary, profit_factor = fit_strategy(
strategy_dictionary, data_local, fitting_inputs_local,
fitting_targets_local)
error_loop = fitting_dictionary['error']
if error_loop < error and fitting_dictionary['n_trades'] != 0:
error = error_loop
strategy_dictionary_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_optimum,
fitting_dictionary_optimum, data_local, toc)
return strategy_dictionary_optimum, fitting_inputs_local, fitting_targets_local, data_local
def random_search(strategy_dictionary_local, n_iterations):
toc = tic()
counter = 0
error = -1e5
while counter < n_iterations:
counter += 1
strategy_dictionary_local = randomise_dictionary_inputs(strategy_dictionary_local)
fitting_dictionary, data_to_predict, profit_factor = fit_strategy(strategy_dictionary_local)
error_loop = fitting_dictionary['error']
if error_loop > error and fitting_dictionary['n_trades'] != 0:
error = error_loop
strategy_dictionary_local_optimum = strategy_dictionary_local
fitting_dictionary_optimum = fitting_dictionary
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary_local_optimum, fitting_dictionary_optimum, data_to_predict, toc)
return strategy_dictionary_local_optimum
def tensorflow_fitting(strategy_dictionary_local):
toc = tic()
data_local, data_2 = import_data(strategy_dictionary_local)
fitting_inputs_local, continuous_targets, classification_targets = input_processing(
data_local, data_2, strategy_dictionary)
if strategy_dictionary_local['regression_mode'] == 'classification':
fitting_targets_local = classification_targets
elif strategy_dictionary_local['regression_mode'] == 'regression':
fitting_targets_local = continuous_targets
fitting_inputs_local, strategy_dictionary_local = preprocessing_inputs(
strategy_dictionary_local, fitting_inputs_local)
fitting_dictionary, error_loop, profit_factor = fit_tensorflow(
strategy_dictionary_local, data_local, fitting_inputs_local,
fitting_targets_local)
underlined_output('Best strategy fit')
output_strategy_results(strategy_dictionary, fitting_dictionary,
data_local, toc)
return strategy_dictionary, data_local, fitting_inputs_local, fitting_targets_local
'ticker_1': 'USDT_BTC',
'scraper_currency_1': 'BTC',
'candle_size': 300,
'n_days': 10,
'offset': 0,
'bid_ask_spread': 0.0007,
'transaction_fee': 0.0025,
'train_test_validation_ratios': [0.5, 0.2, 0.3],
'output_flag': True,
'plot_flag': False,
'plot_last': True,
'ml_iterations': 15,
'target_score': 'n_steps',
'web_flag': True,
'filename1': "USDT_BTC.csv",
'regression_mode': 'regression',
'momentum_compare': True,
}
search_iterations = 15
time_iterations = 30
strategy_dictionary, fitting_inputs, fitting_targets, data_to_predict = fit_time_scale(
strategy_dictionary, search_iterations, time_iterations)
underlined_output('Offset validation')
offsets = np.linspace(0, 300, 5)
offset_scan_validation(strategy_dictionary, data_to_predict,
fitting_inputs, fitting_targets, offsets)
search_iterations = 15
time_iterations = 10
if strategy_dictionary['fit_time']:
strategy_dictionary, fitting_dictionary, fitting_inputs, fitting_targets, data_to_predict = fit_time_scale(
strategy_dictionary, search_iterations, time_iterations, toc)
else:
strategy_dictionary, fitting_dictionary, fitting_inputs, fitting_targets, data_to_predict, test_profit \
= random_search(
strategy_dictionary,
search_iterations,
toc)
underlined_output('Best strategy fit')
if strategy_dictionary['plot_last']:
strategy_dictionary['plot_flag'] = True
output_strategy_results(strategy_dictionary,
fitting_dictionary,
data_to_predict,
toc,
momentum_dict=simple_momentum_comparison(
data_to_predict, strategy_dictionary,
fitting_dictionary))
underlined_output('Offset validation')
offsets = np.linspace(0, 300, 5)
'trading_currencies': ['ETH', 'BTC'],
'ticker_1': 'BTC_ETH',
'ticker_2': 'USDT_BTC',
'candle_size': 1800,
'n_days': 20,
'offset': 0,
'bid_ask_spread': 0.001,
'transaction_fee': 0.0025,
'train_test_ratio': 0.5,
'output_flag': True,
'plot_flag': False,
'ml_iterations': 100,
'target_score': 'idealstrategy',
'windows': [1, 5, 10, 50, 100],
'web_flag': False,
'filename1': "USDT_BTC.csv",
'filename2': "BTC_ETH.csv"
}
search_iterations = 50
underlined_output('Searching strategy')
strategy_dictionary = random_search(strategy_dictionary, search_iterations)
underlined_output('Offset validation')
offsets = np.linspace(0, 300, 5)
offset_scan_validation(strategy_dictionary, offsets)
print strategy_dictionary
