def _get_bar_features(self, date_time: pd.Timestamp,
list_bars: list) -> list:
"""
Calculate inter-bar features: lambdas, entropies, avg_tick_size, vwap
:param date_time: (pd.Timestamp) when bar was formed
:param list_bars: (list) of previously formed bars
:return: (list) of inter-bar features
"""
features = [date_time]
# Tick rule sum, avg tick size, VWAP
features.append(get_avg_tick_size(self.trade_size))
features.append(sum(self.tick_rule))
features.append(vwap(self.dollar_size, self.trade_size))
# Lambdas
features.extend(
get_trades_based_kyle_lambda(self.price_diff, self.trade_size,
self.tick_rule)) # Kyle lambda
features.extend(
get_trades_based_amihud_lambda(self.log_ret,
self.dollar_size)) # Amihud lambda
features.extend(
get_trades_based_hasbrouck_lambda(
self.log_ret, self.dollar_size,
self.tick_rule)) # Hasbrouck lambda
# Entropy features
encoded_tick_rule_message = encode_tick_rule_array(self.tick_rule)
features.append(get_shannon_entropy(encoded_tick_rule_message))
features.append(get_plug_in_entropy(encoded_tick_rule_message))
features.append(get_lempel_ziv_entropy(encoded_tick_rule_message))
features.append(get_konto_entropy(encoded_tick_rule_message))
if self.volume_encoding is not None:
message = encode_array(self.trade_size, self.volume_encoding)
features.append(get_shannon_entropy(message))
features.append(get_plug_in_entropy(message))
features.append(get_lempel_ziv_entropy(message))
features.append(get_konto_entropy(message))
if self.pct_encoding is not None:
message = encode_array(self.log_ret, self.pct_encoding)
features.append(get_shannon_entropy(message))
features.append(get_plug_in_entropy(message))
features.append(get_lempel_ziv_entropy(message))
features.append(get_konto_entropy(message))
list_bars.append(features)
def test_entropy_calculations(self):
"""
Test entropy functions
"""
message = '11100001'
message_array = [1, 1, 1, 0, 0, 0, 0, 1]
shannon = get_shannon_entropy(message)
plug_in = get_plug_in_entropy(message, word_length=1)
plug_in_arr = get_plug_in_entropy(message_array, word_length=1)
lempel = get_lempel_ziv_entropy(message)
konto = get_konto_entropy(message)
self.assertEqual(plug_in, plug_in_arr)
self.assertAlmostEqual(shannon, 1.0, delta=1e-3)
self.assertAlmostEqual(lempel, 0.625, delta=1e-3)
self.assertAlmostEqual(plug_in, 0.985, delta=1e-3)
self.assertAlmostEqual(konto, 0.9682, delta=1e-3)
# Konto entropy boundary conditions
konto_2 = get_konto_entropy(message, 2)
_match_length('1101111', 2, 3)
self.assertAlmostEqual(konto_2, 0.8453, delta=1e-4)
self.assertEqual(get_konto_entropy('a'), 0) # one-character message entropy = 0