def generate_signals(a_timer: PipelineTimer,
arb_id_dict: dict,
signal_pickle_filename: str,
normalize_strategy,
force=False):
if path.isfile(signal_pickle_filename):
if force:
# Remove any existing pickled Signal dictionary and create one.
remove(signal_pickle_filename)
else:
print(
"\nSignal generation already completed and forcing is turned off. Using pickled data..."
)
return load(open(signal_pickle_filename, "rb"))
a_timer.start_function_time()
signal_dict = {}
for k, arb_id in arb_id_dict.items():
if not arb_id.static:
for token in arb_id.tokenization:
a_timer.start_iteration_time()
signal = Signal(k, token[0], token[1])
# Convert the binary ndarray to a list of string representations of each row
temp1 = [
''.join(str(x) for x in row)
for row in arb_id.boolean_matrix[:, token[0]:token[1] + 1]
]
temp2 = zeros((temp1.__len__(), 1), dtype=uint64)
# convert each string representation to int
for i, row in enumerate(temp1):
temp2[i] = int(row, 2)
# create an unsigned integer pandas.Series using the time index from this Arb ID's original data.
signal.time_series = Series(temp2[:, 0],
index=arb_id.original_data.index,
dtype=float64)
# Normalize the signal and update its meta-data
signal.normalize_and_set_metadata(normalize_strategy)
# add this signal to the signal dictionary which is keyed by Arbitration ID
if k in signal_dict:
signal_dict[k][(arb_id.id, signal.start_index,
signal.stop_index)] = signal
else:
signal_dict[k] = {
(arb_id.id, signal.start_index, signal.stop_index):
signal
}
a_timer.set_token_to_signal()
a_timer.set_signal_generation()
return signal_dict
