def x_tick_labels(self):
arr = np.mod(self.index_to_intday, 10000)
is_year_firsts = (arr - np_shift(arr, 1, fill_value=0)) < 0
arr = np.mod(self.index_to_intday, 100)
is_month_firsts = (arr - np_shift(arr, 1, fill_value=0)) < 0
yposes = self.xs[is_year_firsts]
mposes = self.xs[is_month_firsts]
self.time_pos_info = yposes, mposes
return yposes, mposes
def jump_value(df):
# Return 0, negative value, positive value
high = df.high.values
low = df.low.values
prehigh = np_shift(high, 1, 0)
prelow = np_shift(low, 1, 0)
jump_up = (low - prehigh)
jump_up[jump_up <= 0] = 0
jump_down = (high - prelow)
jump_down[jump_down >= 0] = 0
return jump_up + jump_down
def day_compare_features(df):
o, c, h, l = KlineFeatures.ochl_features(df)
po = np_shift(o, 1, 0)
pc = np_shift(c, 1, 0)
ph = np_shift(h, 1, 0)
pl = np_shift(l, 1, 0)
hh = np.maximum(h, ph)
ll = np.minimum(l, pl)
rr = hh - ll + sys.float_info.epsilon
f1 = (o - ll) / rr
f2 = (c - ll) / rr
f3 = (po - ll) / rr
f4 = (pc - ll) / rr
f5 = (h - ll) / rr
f6 = (l - ll) / rr
f7 = (ph - ll) / rr
f8 = (pl - ll) / rr
return f1, f2, f3, f4, f5, f6, f7, f8
def even_count(arr):
is_evens = (arr == np_shift(arr, 1, 0))
counts = ArrayIndicator.consecutive_count_of_True(is_evens)
return counts
def drop_count(arr):
is_drops = arr - np_shift(arr, 1, 0) < 0
counts = ArrayIndicator.consecutive_count_of_True(is_drops)
return counts
def rise_count(arr):
is_rises = arr - np_shift(arr, 1, 0) > 0
counts = ArrayIndicator.consecutive_count_of_True(is_rises)
return counts
def func(df):
val = cls.day_compare_features(df)
return [np_shift(item, n, 0) for item in val]
def is_close_up(df):
close_ = df.close.values
close_pre = np_shift(close_, 1, 0)
return close_ - close_pre > 0
def even_count_of_previous_days(df):
arr = df.close.values
is_even = (arr == np_shift(arr, 1, 0))
count = ArrayIndicator.consecutive_count_of_True(is_even)
return np_shift(count, 1, 0)
def down_count_of_previous_days(df):
arr = df.close.values
is_up = arr - np_shift(arr, 1, 0) < 0
count = ArrayIndicator.consecutive_count_of_True(is_up)
return np_shift(count, 1, 0)
