async def _build_train_data(self, frame_type: FrameType,
n: int, max_error: float = 0.01):
"""
从最近的符合条件的日期开始，遍历股票，提取特征和标签，生成数据集。
Args:
n: 需要采样的样本数
Returns:
"""
watch_win = 5
max_curve_len = 5
max_ma_win = 20
# y_stop = arrow.get('2020-7-24').date()
y_stop = tf.floor(arrow.now(tz=cfg.tz), frame_type)
y_start = tf.shift(y_stop, -watch_win + 1, frame_type)
x_stop = tf.shift(y_start, -1, frame_type)
x_start = tf.shift(x_stop, -(max_curve_len + max_ma_win - 1), frame_type)
data = []
while len(data) < n:
for code in Securities().choose(['stock']):
#for code in ['000601.XSHE']:
try:
sec = Security(code)
x_bars = await sec.load_bars(x_start, x_stop, FrameType.DAY)
y_bars = await sec.load_bars(y_start, y_stop, FrameType.DAY)
# [a, b, axis] * 3
x = self.extract_features(x_bars, max_error)
if len(x) == 0: continue
y = np.max(y_bars['close']) / x_bars[-1]['close'] - 1
if np.isnan(y): continue
feature = [code, tf.date2int(x_stop)]
feature.extend(x)
data.append(feature)
except Exception as e:
logger.warning("Failed to extract features for %s (%s)",
code,
x_stop)
logger.exception(e)
if len(data) >= n:
break
if len(data) % 500 == 0:
logger.info("got %s records.", len(data))
y_stop = tf.day_shift(y_stop, -1)
y_start = tf.day_shift(y_start, -1)
x_stop = tf.day_shift(y_start, -1)
x_start = tf.day_shift(x_start, -1)
return data
async def test_buy_limit_events(self):
end = arrow.get('2020-8-7').date()
start = tf.day_shift(end, -9)
sec = Security('603390.XSHG')
bars = await sec.load_bars(start, end, FrameType.DAY)
count, indices = count_buy_limit_event(sec, bars)
self.assertEqual(count, 1)
self.assertEqual(
arrow.get('2020-7-28').date(), bars['frame'][indices[0]])
sec = Security('000070.XSHE')
start = tf.day_shift(end, -29)
bars = await sec.load_bars(start, end, FrameType.DAY)
count_buy_limit_event(sec, bars)
async def predict(self, code, x_end_date: datetime.date, max_error: float = 0.01):
sec = Security(code)
start = tf.day_shift(x_end_date, -29)
bars = await sec.load_bars(start, x_end_date, FrameType.DAY)
features = self.extract_features(bars, max_error)
if len(features) == 0:
logger.warning("cannot extract features from %s(%s)", code, x_end_date)
else:
return self.model.predict([features])
async def test_cross(self):
end = arrow.get('2020-7-24').date()
start = tf.day_shift(end, -270)
sec = Security('000035.XSHE')
jlkg = await sec.load_bars(start, end, FrameType.DAY)
ma5 = signal.moving_average(jlkg['close'], 5)
ma250 = signal.moving_average(jlkg['close'], 250)
flag, idx = signal.cross(ma5[-10:], ma250[-10:])
self.assertEqual(flag, -1)
self.assertEqual(idx, 8)
def parse_sync_params(
frame: Union[str, Frame],
cat: List[str] = None,
start: Union[str, datetime.date] = None,
stop: Union[str, Frame] = None,
delay: int = 0,
include: str = "",
exclude: str = "",
) -> Tuple:
"""按照[使用手册](usage.md#22-如何同步K线数据)中的规则，解析和补全同步参数。
如果`frame_type`为分钟级，则当`start`指定为`date`类型时，自动更正为对应交易日的起始帧；
当`stop`为`date`类型时，自动更正为对应交易日的最后一帧。
Args:
frame (Union[str, Frame]): frame type to be sync. The word ``frame`` is used
here for easy understand by end user. It actually implies "FrameType".
cat (List[str]): which catetories is about to be synced. Should be one of
['stock', 'index']. Defaults to None.
start (Union[str, datetime.date], optional): [description]. Defaults to None.
stop (Union[str, Frame], optional): [description]. Defaults to None.
delay (int, optional): [description]. Defaults to 5.
include (str, optional): which securities should be included, seperated by
space, for example, "000001.XSHE 000004.XSHE". Defaults to empty string.
exclude (str, optional): which securities should be excluded, seperated by
a space. Defaults to empty string.
Returns:
- codes (List[str]): 待同步证券列表
- frame_type (FrameType):
- start (Frame):
- stop (Frame):
- delay (int):
"""
frame_type = FrameType(frame)
if frame_type in tf.minute_level_frames:
if stop:
stop = arrow.get(stop, tzinfo=cfg.tz)
if stop.hour == 0: # 未指定有效的时间帧，使用当日结束帧
stop = tf.last_min_frame(tf.day_shift(stop.date(), 0),
frame_type)
else:
stop = tf.floor(stop, frame_type)
else:
stop = tf.floor(arrow.now(tz=cfg.tz).datetime, frame_type)
if stop > arrow.now(tz=cfg.tz):
raise ValueError(f"请勿将同步截止时间设置在未来: {stop}")
if start:
start = arrow.get(start, tzinfo=cfg.tz)
if start.hour == 0: # 未指定有效的交易帧，使用当日的起始帧
start = tf.first_min_frame(tf.day_shift(start.date(), 0),
frame_type)
else:
start = tf.floor(start, frame_type)
else:
start = tf.shift(stop, -999, frame_type)
else:
stop = (stop and arrow.get(stop).date()) or arrow.now().date()
if stop == arrow.now().date():
stop = arrow.now(tz=cfg.tz)
stop = tf.floor(stop, frame_type)
start = tf.floor(
(start and arrow.get(start).date()), frame_type) or tf.shift(
stop, -1000, frame_type)
secs = Securities()
codes = secs.choose(cat or [])
exclude = map(lambda x: x, exclude.split(" "))
codes = list(set(codes) - set(exclude))
include = list(filter(lambda x: x, include.split(" ")))
codes.extend(include)
return codes, frame_type, start, stop, int(delay)