def get_rsi(self, symbol: str, date: str, n: int):
""" 计算所有日期的 RSI(n) 序列, 返回给定日期的 RSI(n) """
key = f'rsi-{n}'
if self.computed_memo.contains(symbol, key):
if len(self.computed_memo.get(symbol, key)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
if date in global_data_instance.symbol_to_date_set[symbol]:
offset = global_data_instance.find_date_offset(
symbol, date)
return self.computed_memo.get(symbol, key)[offset]
else:
return np.nan
close_array = global_data_instance.get_array_since_date(
symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
# LC:=REF(CLOSE,1);
lc = ref(close_array, 1)
# RSI:SMA(MAX(CLOSE-LC,0),N1,1)/SMA(ABS(CLOSE-LC),N1,1)*100;
rsi = self.sma(np.maximum(close_array - lc, 0), n, 1) / (
self.sma(np.abs(close_array - lc), n, 1) + 1e-6) * 100 # 避免0÷0
self.computed_memo.set(symbol, key, rsi) # 计算出来后填入缓存
offset = global_data_instance.find_date_offset(symbol, date)
return rsi[offset]
def get_cci(self, symbol: str, date: str, n: int):
""" 计算所有日期的 CCI(n) 序列, 返回给定日期的 CCI(n) """
key = f'cci-{n}'
if self.computed_memo.contains(symbol, key):
if len(self.computed_memo.get(symbol, key)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
if date in global_data_instance.symbol_to_date_set[symbol]:
offset = global_data_instance.find_date_offset(
symbol, date)
return self.computed_memo.get(symbol, key)[offset]
else:
return np.nan
close_array = global_data_instance.get_array_since_date(
symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
high_array = global_data_instance.get_array_since_date(
symbol, OHLCV.HIGH, global_data_instance.START_DOWNLOAD_DATE)
low_array = global_data_instance.get_array_since_date(
symbol, OHLCV.LOW, global_data_instance.START_DOWNLOAD_DATE)
# TYP:=(HIGH+LOW+CLOSE)/3;
# CCI:(TYP-MA(TYP,N))/(0.015*AVEDEV(TYP,N));
typ = (high_array + low_array + close_array) / 3
cci = (typ - self.ma(typ, n)) / (0.015 * self.avedev(typ, n))
self.computed_memo.set(symbol, key, cci) # 计算出来后填入缓存
offset = global_data_instance.find_date_offset(symbol, date)
return cci[offset]
def get_kdj(self, symbol: str, date: str, n: int, m: int):
""" 计算所有日期的 KDJ 序列, 返回给定日期的 KDJ 的J值 """
key = f'kdj-{n}-{m}'
if self.computed_memo.contains(symbol, key):
if len(self.computed_memo.get(symbol, key)) == len(global_data_instance.symbol_to_date_list[symbol]): # 并且array长度相等(数据存在且正确)
if date in global_data_instance.symbol_to_date_set[symbol]:
offset = global_data_instance.find_date_offset(symbol, date)
return self.computed_memo.get(symbol, key)[offset]
else:
return np.nan
close_array = global_data_instance.get_array_since_date(symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
high_array = global_data_instance.get_array_since_date(symbol, OHLCV.HIGH, global_data_instance.START_DOWNLOAD_DATE)
low_array = global_data_instance.get_array_since_date(symbol, OHLCV.LOW, global_data_instance.START_DOWNLOAD_DATE)
# 获取LLV(N)
key = f'llv-{n}'
llv = None
if self.computed_memo.contains(symbol, key): # 已有计算
if len(self.computed_memo.get(symbol, key)) == len(global_data_instance.symbol_to_date_list[symbol]): # 并且array长度相等(数据存在且正确)
llv = self.computed_memo.get(symbol, key)
if llv is None:
llv = self.llv(low_array, n)
self.computed_memo.set(symbol, key, llv) # 计算出来后填入缓存
# 获取HHV(N)
key = f'hhv-{n}'
hhv = None
if self.computed_memo.contains(symbol, key): # 已有计算
if len(self.computed_memo.get(symbol, key)) == len(global_data_instance.symbol_to_date_list[symbol]): # 并且array长度相等(数据存在且正确)
hhv = self.computed_memo.get(symbol, key)
if hhv is None:
hhv = self.hhv(high_array, n)
self.computed_memo.set(symbol, key, hhv) # 计算出来后填入缓存
# RSV:=(CLOSE-LLV(LOW,N))/(HHV(HIGH,N)-LLV(LOW,N))*100;
rsv: np.ndarray = (close_array - llv) / (hhv - llv + 0.00000001) * 100 # 避免除以0
if np.isnan(rsv[0]): # 若第一天停牌 则hhv-llv等于0 相除之后会变成nan 导致之后的计算全部错误
rsv[0] = 0
k = self.sma(rsv, m, 1)
d = self.sma(k, m, 1)
j = 3 * k - 2 * d
# 计算KDJ需要较多运算 不能直接读取(要算SMA) 所以把结果暂时存下来
key = f'kdj-{n}-{m}'
self.computed_memo.set(symbol, key, j)
offset = global_data_instance.find_date_offset(symbol, date)
return j[offset]
def get_dmi(self, symbol: str, date: str, n: int):
""" 计算所有日期的 PDI、MDI 序列, 返回给定日期的 (PDI, MDI) 值 """
key_pdi = f'pdi-{n}'
pdi = None
if self.computed_memo.contains(symbol, key_pdi): # 已有计算
if len(self.computed_memo.get(symbol, key_pdi)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
pdi = self.computed_memo.get(symbol, key_pdi)
key_mdi = f'mdi-{n}'
mdi = None
if self.computed_memo.contains(symbol, key_mdi): # 已有计算
if len(self.computed_memo.get(symbol, key_mdi)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
mdi = self.computed_memo.get(symbol, key_mdi)
if (pdi is not None) and (mdi is not None):
if date in global_data_instance.symbol_to_date_set[symbol]:
offset = global_data_instance.find_date_offset(symbol, date)
return (self.computed_memo.get(symbol, key_pdi)[offset],
self.computed_memo.get(symbol, key_mdi)[offset])
close_array = global_data_instance.get_array_since_date(
symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
high_array = global_data_instance.get_array_since_date(
symbol, OHLCV.HIGH, global_data_instance.START_DOWNLOAD_DATE)
low_array = global_data_instance.get_array_since_date(
symbol, OHLCV.LOW, global_data_instance.START_DOWNLOAD_DATE)
mtr = calc_mtr(high_array, low_array, close_array, n)
hd: np.ndarray = np.nan_to_num(high_array - ref(high_array, 1))
ld: np.ndarray = np.nan_to_num(ref(low_array, 1) - low_array)
dmp = sum_recent(np.where(np.logical_and(hd > 0, hd > ld), hd, 0), n)
dmm = sum_recent(np.where(np.logical_and(ld > 0, ld > hd), ld, 0), n)
pdi = dmp * 100 / mtr
mdi = dmm * 100 / mtr
self.computed_memo.set(symbol, key_pdi, pdi) # 计算出来后填入缓存
self.computed_memo.set(symbol, key_mdi, mdi) # 计算出来后填入缓存
offset = global_data_instance.find_date_offset(symbol, date)
return (pdi[offset], mdi[offset])
def get_ma(self, symbol: str, date: str, n: int) -> float:
""" 计算所有日期的 N日均线 的序列（若有缓存则无需计算），返回给定日期的 N日均线 的值 """
key = f'ma-{n}'
if self.computed_memo.contains(symbol, key): # 已有计算
if len(self.computed_memo.get(symbol, key)) == len(global_data_instance.symbol_to_date_list[symbol]): # 并且array长度相等(数据存在且正确)
if date in global_data_instance.symbol_to_date_set[symbol]:
offset = global_data_instance.find_date_offset(symbol, date)
return self.computed_memo.get(symbol, key)[offset]
else:
return np.nan
close_array = global_data_instance.get_array_since_date(symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
# 计算MA
if len(close_array) > n: # 避免只有50天数据却计算了MA90的问题 否则求卷积后提取时会有问题
ma_array = self.ma(close_array, n)
self.computed_memo.set(symbol, key, ma_array) # 计算出来后填入缓存
offset = global_data_instance.find_date_offset(symbol, date)
return ma_array[offset]
else:
return np.nan
def get_macd(self, symbol: str, date: str, short: int, long: int,
mid: int):
""" 计算所有日期的MACD序列, 返回给定日期的MACD """
close_array = global_data_instance.get_array_since_date(
symbol, OHLCV.CLOSE, global_data_instance.START_DOWNLOAD_DATE)
# 获取EMA(CLOSE,SHORT)
key = f'ema-{short}'
ema_short = None
if self.computed_memo.contains(symbol, key): # 已有计算
if len(self.computed_memo.get(symbol, key)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
ema_short = self.computed_memo.get(symbol, key)
if ema_short is None:
ema_short = self.ema(close_array, short)
self.computed_memo.set(symbol, key, ema_short) # 计算出来后填入缓存
# 获取EMA(CLOSE,LONG)
key = f'ema-{long}'
ema_long = None
if self.computed_memo.contains(symbol, key): # 已有计算
if len(self.computed_memo.get(symbol, key)) == len(
global_data_instance.symbol_to_date_list[symbol]
): # 并且array长度相等(数据存在且正确)
ema_long = self.computed_memo.get(symbol, key)
if ema_long is None:
ema_long = self.ema(close_array, long)
self.computed_memo.set(symbol, key, ema_long) # 计算出来后填入缓存
# DIF:EMA(CLOSE,SHORT)-EMA(CLOSE,LONG);
dif: np.ndarray = ema_short - ema_long
# DEA:EMA(DIF,MID);
dea: np.ndarray = self.ema(dif, mid)
# MACD:(DIF-DEA)*2;
macd: np.ndarray = (dif - dea) * 2
key = f'macd-{short}-{long}-{mid}'
self.computed_memo.set(symbol, key, macd)
offset = global_data_instance.find_date_offset(symbol, date)
return macd[offset]