def run():
sound_state = False
error = False
while (True):
op = t.general_menu(error=error)
error = False
if (op == "1"): a.run()
elif (op == "2"): sound_state = pause(sound_state)
elif (op == "3"): instructions()
elif (op == "4"): t.credits()
elif (op == "5"): break
else: error = True
def saw_eval_base(individual, stocks, bundle_name, train_start, train_end,
capital_base, trade_freq, social_media, **kwargs):
# convert individual weights into weights for the stocks
# if debug: print("SAW EVAL BASE")
w = OrderedDict()
i = 0
for s in stocks:
w[s] = {"p": individual[i], "n": individual[i + 1]}
i = i + 2
algo = TradingSignalAlgorithm(
verbose=False,
grp="DALIO",
subgrp="ALL_WEATHER",
collect_before_trading=False,
history=500,
rebalance_freq=trade_freq,
trading_signal=saw_ga_trading_fn,
initial_weights=[0.3, 0.4, 0.15, 0.075, 0.075],
normalise_weights=True,
**{
"weights": w,
"social_media": social_media
})
_, algo_results = run("SAW-GA",
algo,
bundle_name,
train_start,
train_end,
capital_base,
analyze=False)
return algo_results
def smpt_eval_base(individual,
stocks,
bundle_name,
train_start,
train_end,
capital_base,
trade_freq,
social_media,
objective="max_sharpe",
**kwargs):
# if debug: print("SMPT EVAL BASE")
w = list(individual)
algo = OptAlgorithm(verbose=False,
grp="DALIO",
subgrp="ALL_WEATHER",
collect_before_trading=False,
history=500,
rebalance_freq=trade_freq,
mpt_adjustment=smpt_ga_trading_fn,
objective=objective,
**{
"weights": w,
"social_media": social_media
})
_, algo_results = run("SMPT-GA",
algo,
bundle_name,
train_start,
train_end,
capital_base,
analyze=False)
return algo_results
#!/usr/bin/env python
import algorithms
import os
#filename = 'iris.data' # expects csv
filename = 'gtd.data' # expects csv
filename = os.curdir + "/" + filename
# pass fully qualified file name to run()
# TODO: use 'argparse' module to take this from command line
algorithms.run(filename, 'perceptron', {'testSplit':0.2, 'learningRate':0.005, \
'maxIterations':10000})
#!/usr/bin/env python
import algorithms
import os
#filename = 'iris.data' # expects csv
filename = 'gtd.data' # expects csv
filename = os.curdir + "/" + filename
# pass fully qualified file name to run()
# TODO: use 'argparse' module to take this from command line
algorithms.run(filename, 'perceptron', {'testSplit':0.2, 'learningRate':0.005, \
'maxIterations':10000})
def compareResults(base_name="SAW_GA_MAX_RET",
opt_type="saw",
social_media=None,
bundle_name="",
train_start=None,
test_start=None,
test_end=None,
stocks=None,
trade_freq='weekly',
capital_base=1000000,
history=500,
prefix="",
**kwargs):
all_ga = [] # [bm_all_weather]
test_ga = [] # [bm_aw_test]
for file in tqdm(glob.glob(f"{base_name}_p*.pickle")):
file = file.replace("\\", "/")
m = re.search(f'{base_name}_p(\d*)_g(\d*)_s(\d*).*', file)
# print(file)
pop = m.group(1)
gen = m.group(2)
seed = m.group(3)
with open(f"{file}", "rb+") as f:
top10_max_ret = pickle.load(f)
if opt_type == "saw":
best_max_ret = top10_max_ret[0]
w_max_ret = OrderedDict()
i = 0
for s in stocks:
w_max_ret[s] = {
"p": best_max_ret[i],
"n": best_max_ret[i + 1]
}
i = i + 2
algo = TradingSignalAlgorithm(
verbose=False,
grp="DALIO",
subgrp="ALL_WEATHER",
collect_before_trading=False,
history=history,
rebalance_freq=trade_freq,
trading_signal=saw_ga_trading_fn,
initial_weights=[0.3, 0.4, 0.15, 0.075, 0.075],
normalise_weights=True,
**{
"weights": w_max_ret,
"social_media": social_media
})
else:
w_max_ret = top10_max_ret[0]
algo = OptAlgorithm(verbose=False,
grp="DALIO",
subgrp="ALL_WEATHER",
collect_before_trading=False,
history=history,
rebalance_freq=trade_freq,
mpt_adjustment=smpt_ga_trading_fn,
**{
"weights": w_max_ret,
"social_media": social_media
})
saw_ga_test = run(f"{prefix}{pop}-{gen}-{seed}",
algo,
bundle_name,
test_start,
test_end,
capital_base,
analyze=False)
saw_ga_all = run(f"{prefix}{pop}-{gen}-{seed}",
algo,
bundle_name,
train_start,
test_end,
capital_base,
analyze=False)
test_ga.append(saw_ga_test)
all_ga.append(saw_ga_all)
return test_ga, all_ga