def inclination(self, data_i):
"""
標準偏差の傾きを調べる。ボラティリティの広がりをパターンに分ける
:param data_i: int
:return: int: 定数、傾きのパターン
"""
pre_data = self._data.loc[data_i - self._last_data_num + 1:data_i,
'sigma'].values
min_price = np.amin(pre_data)
t = pre_data - np.full_like(a=pre_data, fill_value=min_price)
x = np.arange(start=0,
step=self._inclination_alpha,
stop=self._inclination_alpha * len(t))
inclination = functions.linear_regression(
x=x, t=t, basic_function=functions.Polynomial(dim=2))[1]
if self.POSITIVE_INCLINATION < inclination:
inclination_pattern = self.HYPER_EXPANSION
elif (self.POSITIVE_MIDDLE_INCLINATION <
inclination) and (inclination <= self.POSITIVE_INCLINATION):
inclination_pattern = self.EXPANSION
elif (self.NEGATIVE_MIDDLE_INCLINATION <= inclination) and (
inclination <= self.POSITIVE_MIDDLE_INCLINATION):
inclination_pattern = self.FLAT
elif (self.NEGATIVE_INCLINATION <= inclination) and (
inclination < self.NEGATIVE_MIDDLE_INCLINATION):
inclination_pattern = self.SQUEEZE
elif inclination < self.NEGATIVE_INCLINATION:
inclination_pattern = self.HYPER_SQUEEZE
else:
inclination_pattern = None
return inclination_pattern
def __inclination(self):
min_sigma = np.amin(self.__recent_sigma)
# 最小値との差分だけの行列を作る
t = self.__recent_sigma - np.full_like(a=self.__recent_sigma,
fill_value=min_sigma)
t = t * 1000
x = np.arange(start=0,
step=self.__inclination_alpha,
stop=self.__inclination_alpha * len(t))
# 直線(１次多項式)の線形回帰
# その傾きを取り出す
inclination = linear_regression(x=x,
t=t,
basic_function=Polynomial(dim=2))[1]
if self.POSITIVE_INCLINATION < inclination:
inclination_pattern = self.HYPER_EXPANSION
elif (self.POSITIVE_MIDDLE_INCLINATION <
inclination) and (inclination <= self.POSITIVE_INCLINATION):
inclination_pattern = self.EXPANSION
elif (self.NEGATIVE_MIDDLE_INCLINATION <= inclination) and (
inclination <= self.POSITIVE_MIDDLE_INCLINATION):
inclination_pattern = self.FLAT
elif (self.NEGATIVE_INCLINATION <= inclination) and (
inclination < self.NEGATIVE_MIDDLE_INCLINATION):
inclination_pattern = self.SQUEEZE
elif inclination < self.NEGATIVE_INCLINATION:
inclination_pattern = self.HYPER_SQUEEZE
else:
raise TypeError('inclination is None')
return inclination_pattern
def test_linear_regression(self):
x = np.asarray([0, 20, 40])
t = np.asarray([100, 120, 140])
polynomial = functions.Polynomial(dim=2)
w = functions.linear_regression(x=x, t=t, basic_function=polynomial)
self.assertEqual(100, round(w[0]))
self.assertEqual(1, round(w[1]))
print(round(w[0]), w[0])
print(round(w[1]), w[1])
def inclination(self, data_i):
"""
標準偏差の傾きを調べる。ボラティリティの広がりをパターンに分ける
:param data_i:
:return:
"""
pre_data = self.__data.loc[data_i - self.__stock_term + 1:data_i,
self.__target].values
min_price = np.amin(pre_data)
t = pre_data - np.full_like(a=pre_data, fill_value=min_price)
t = t * 1000
x = np.arange(start=0,
step=self.__inclination_alpha,
stop=self.__inclination_alpha * len(t))
inclination = functions.linear_regression(
x=x, t=t, basic_function=functions.Polynomial(dim=2))[1]
self.__inclination_list.append(inclination)
self.inclination_pattern(inclination=inclination)