def short(self, cols, strict=True, inplace=False, sig_name=None):
rdf = DataFrame()
for i in range(1, len(cols)):
name = cols[i - 1] + '_' + cols[i]
rdf[name] = StockSeries(self.df[cols[i - 1]]).lower_than(
self.df[cols[i]], strict)
rdf[name] = rdf[name].map(lambda x: True if x == 1 else False)
signal = rdf.all(axis=1).map(lambda x: 1 if x is True else 0)
if inplace:
if sig_name is None:
sig_name = 'short_' + '_'.join(cols)
self.df[sig_name] = signal
else:
return StockSeries(signal)
def decrease(self, cols, strict=True, inplace=False, sig_name=None):
if isinstance(cols, str):
signal = StockSeries(self.df[cols]).increase(strict)
else:
rdf = DataFrame()
for col in cols:
rdf[col] = StockSeries(self.df[col]).wave(strict).map(
lambda x: True if x < 0 else False)
signal = rdf.all(axis=1).map(lambda x: 1 if x is True else 0)
if inplace:
if sig_name is None:
sig_name = 'decrease_' + '_'.join(cols)
self.df[sig_name] = signal
else:
return StockSeries(signal)
def llv(self, n=5, col='close', inplace=False):
llv_s = self.df[col].rolling(center=False, window=n).min()
col_name = self._get_name(col + '_LLV_' + str(n))
if inplace:
self.df[col_name] = llv_s
else:
return StockSeries(llv_s, col_name)
def hhv(self, n=5, col='close', inplace=False):
hhv_s = self.df[col].rolling(center=False, window=n).max()
col_name = self._get_name(col + '_HHV_' + str(n))
if inplace:
self.df[col_name] = hhv_s
else:
return StockSeries(hhv_s, col_name)
def expma(self, n=5, col='close', inplace=False):
expma_s = round(self.df[col].ewm(span=n, adjust=False).mean(), 2)
col_name = self._get_name(col + '_EXPMA_' + str(n))
if inplace:
self.df[col_name] = expma_s
else:
return StockSeries(expma_s, col_name)
def ma(self, n=5, col='close', inplace=False):
ma_s = round(self.df[col].rolling(center=False, window=n).mean(), 2)
col_name = self._get_name(col + '_MA_' + str(n))
if inplace:
self.df[col_name] = ma_s
else:
return StockSeries(ma_s, col_name)
def __getitem__(self, x):
if isinstance(x, str):
return StockSeries(self.df[x], x)
elif isinstance(x, list):
return StockDataFrame(self.df[x])
else:
raise OSError('Error: Unexpected index for StockDataFrame.')
def rsi(self, n=6, inplace=False):
rsi_n = talib.RSI(tdf['close'].values, timeperiod=n)
rsi_s = round(pd.Series(rsi_n, index=self.df.index).fillna(0), 1)
col_name = 'RSI_' + str(n)
if inplace:
self.df[col_name] = rsi_s
else:
return StockSeries(rsi_s, col_name)
def wr(self, period=14, inplace=False):
high = self.df['high'].values
low = self.df['low'].values
close = self.df['close'].values
wr_n = talib.WILLR(high, low, close, period) * -1
wr_s = pd.Series(wr_n, index=self.df.index)
col_name = 'WR_' + str(period)
if inplace:
self.df[col_name] = wr_s
else:
return StockSeries(wr_s, col_name)
def bbi(self, nums=[3, 6, 12, 24], col='close', inplace=False):
ma = DataFrame()
for n in nums:
ma[n] = round(self.ma(n, col), 2)
bbi_s = round(ma[nums].apply(sum, axis=1) / 4, 2)
col_name = self._get_name(col + '_BBI_' + str(n))
if inplace:
self.df[col_name] = bbi_s
else:
return StockSeries(bbi_s, col_name)
def _last_min_simple(self, n=3, col='close', inplace=False):
# n: 平滑天数，小于n天的波动被忽略
s = self.df[col]
l = len(s)
keep_wave_days = [0 for _ in range(l)]
last_min_s = self.df[col].copy(deep=True)
up = True
for i in range(1, l):
if s[i - 1] < s[i]:
if up:
keep_wave_days[i] = keep_wave_days[i - 1] + 1
else:
keep_wave_days[i] = 1
up = True
else:
if up:
keep_wave_days[i] = -1
up = False
else:
keep_wave_days[i] = keep_wave_days[i - 1] - 1
for i in range(1, l):
if keep_wave_days[i] == 1:
if i == 1:
last_min_s[i] = last_min_s[i - 1]
#NOTE:前一天那轮的连续下跌时间较短，讨论
elif -keep_wave_days[i - 1] < n:
last_max_idx = i - 1 - abs(keep_wave_days[i - 1])
last_min_idx = last_max_idx - abs(
keep_wave_days[last_max_idx])
# 考虑可能的越界，应该不存在，s[0] = 0
#if last_min_idx < 0:
# print(i)
# last_min_s[i] = s[i-1]
#NOTE: 【前一天(低点)】低于【上轮的低点】，认新低
if s[i - 1] < s[last_min_idx]:
last_min_s[i] = s[i - 1]
#NOTE: 上一轮的低点更低，前一天认为是短时波动，抹平
else:
last_min_s[i] = last_min_s[last_max_idx]
#NOTE:前一天那轮的连续下跌超过n天，直接认定前一天为新低
else:
last_min_s[i] = s[i - 1]
else:
last_min_s[i] = last_min_s[i - 1]
if inplace:
self.df[self._get_name(col + '_LASTMIN_' + str(n))] = last_min_s
else:
return StockSeries(last_min_s)
def last_max(self, n=5, col='close', inplace=False):
# n: 平滑天数，小于n天的波动被忽略
s = self.df[col]
l = len(s)
keep_wave_days = [0 for _ in range(l)]
last_max_s = self.df[col].copy(deep=True)
up = True
for i in range(1, l):
if s[i - 1] < s[i]:
if up:
keep_wave_days[i] = keep_wave_days[i - 1] + 1
else:
keep_wave_days[i] = 1
up = True
else:
if up:
keep_wave_days[i] = -1
up = False
else:
keep_wave_days[i] = keep_wave_days[i - 1] - 1
for i in range(1, l):
if keep_wave_days[i] == -1:
if i == 1:
last_max_s[i] = last_max_s[i - 1]
#NOTE:前一天那轮的连续上涨时间较短，讨论
elif keep_wave_days[i - 1] < n:
last_min_idx = i - 1 - abs(keep_wave_days[i - 1])
last_max_idx = last_min_idx - abs(
keep_wave_days[last_min_idx])
# 考虑可能的越界
if last_max_idx < 0:
print(i)
last_max_s[i] = s[i - 1]
#NOTE: 【前一天(高点)】高于【上轮的高点】，讨论
if s[i - 1] > s[last_max_idx]:
#last_max_s[i] = s[i-1]
#NOTE: 看三段的长度是否超过n，如果三短时间很短，并且涨多跌少，讨论
if i-1 - last_max_idx + abs(keep_wave_days[last_max_idx]) < n and \
abs(keep_wave_days[i-1]+keep_wave_days[last_max_idx]) > abs(keep_wave_days[last_min_idx]):
#NOTE: 还没涨超【上轮高点】的last_max_s，可能是震荡或慢涨没见顶，先保持之前的last_max_s
if s[i - 1] < last_max_s[last_max_idx]:
last_max_s[i] = last_max_s[last_min_idx]
#NOTE: 涨超了【上轮高点的last_max_s，创新高了直接认
else:
last_max_s[i] = s[i - 1]
#NOTE:超过n了，认为是确立上涨趋势,上轮下跌是短期波动，直接认新高
else:
last_max_s[i] = s[i - 1]
#NOTE: 上一轮的高点更高，前一天认为是短时反抽，抹平
else:
last_max_s[i] = last_max_s[last_min_idx]
#NOTE:前一天那轮的连续上涨超过n天，直接认定前一天为新高
else:
last_max_s[i] = s[i - 1]
else:
last_max_s[i] = last_max_s[i - 1]
col_name = self._get_name(col + '_LASTMAX_' + str(n))
if inplace:
self.df[col_name] = last_max_s
else:
return StockSeries(last_max_s, col_name)
def last_min(self, n=5, col='close', inplace=False):
# n: 平滑天数，小于n天的波动被忽略
s = self.df[col]
l = len(s)
keep_wave_days = [0 for _ in range(l)]
last_min_s = self.df[col].copy(deep=True)
up = True
for i in range(1, l):
if s[i - 1] < s[i]:
if up:
keep_wave_days[i] = keep_wave_days[i - 1] + 1
else:
keep_wave_days[i] = 1
up = True
else:
if up:
keep_wave_days[i] = -1
up = False
else:
keep_wave_days[i] = keep_wave_days[i - 1] - 1
for i in range(1, l):
if keep_wave_days[i] == 1:
if i == 1:
last_min_s[i] = last_min_s[i - 1]
#NOTE:前一天那轮的连续下跌时间较短，讨论
elif -keep_wave_days[i - 1] < n:
last_max_idx = i - 1 - abs(keep_wave_days[i - 1])
last_min_idx = last_max_idx - abs(
keep_wave_days[last_max_idx])
# 考虑可能的越界
if last_min_idx < 0:
print(i)
last_min_s[i] = s[i - 1]
#NOTE: 【前一天(低点)】低于【上轮的低点】，讨论
if s[i - 1] < s[last_min_idx]:
#last_min_s[i] = s[i-1]
#NOTE: 看三段的长度是否超过n，如果三短时间很短，并且跌多涨少，讨论
if i-1 - last_min_idx + abs(keep_wave_days[last_min_idx]) < n and \
abs(keep_wave_days[i-1]+keep_wave_days[last_min_idx]) > keep_wave_days[last_max_idx]:
#NOTE: 还没跌超【上轮低点】的last_min_s，可能是震荡或阴跌没见底，先保持之前的last_min_s
if s[i - 1] > last_min_s[last_min_idx]:
last_min_s[i] = last_min_s[last_max_idx]
#NOTE: 跌超了【上轮低点的last_min_s，之前累积的阴跌太多，认低点
else:
last_min_s[i] = s[i - 1]
#NOTE:超过n了，认为是下跌趋势,上轮上涨是短期波动，直接认新低
else:
last_min_s[i] = s[i - 1]
#NOTE: 上一轮的低点更低，前一天认为是短时波动，抹平
else:
last_min_s[i] = last_min_s[last_max_idx]
#NOTE:前一天那轮的连续下跌超过n天，直接认定前一天为新低
else:
last_min_s[i] = s[i - 1]
else:
last_min_s[i] = last_min_s[i - 1]
col_name = self._get_name(col + '_LASTMIN_' + str(n))
if inplace:
self.df[col_name] = last_min_s
else:
return StockSeries(last_min_s, col_name)