def code_classify(code_list, classify_list):
"""
按照code整理其所属的classify
"""
f = h5py.File(conf.HDF5_FILE_CLASSIFY, 'a')
console.write_head(conf.HDF5_OPERATE_ARRANGE, conf.HDF5_RESOURCE_TUSHARE,
conf.HDF5_OTHER_CODE_CLASSIFY)
# 获取classify列表
code_classify_df = tool.init_empty_df(["date", "code", "classify"])
today_str = tradetime.get_today()
for ctype in classify_list:
for classify_name in f[ctype]:
if f[ctype][classify_name].get(conf.HDF5_CLASSIFY_DS_CODE) is None:
console.write_msg(classify_name + "的code列表不存在")
classify_df = tool.df_from_dataset(f[ctype][classify_name],
conf.HDF5_CLASSIFY_DS_CODE,
None)
for index, row in classify_df.iterrows():
code = row[0].astype(str)
if code in code_list:
code_dict = dict()
code_dict["date"] = today_str
code_dict["code"] = code
code_dict["classify"] = classify_name
code_classify_df = code_classify_df.append(
code_dict, ignore_index=True)
console.write_tail()
f.close()
f_other = h5py.File(conf.HDF5_FILE_OTHER, 'a')
tool.delete_dataset(f_other, conf.HDF5_OTHER_CODE_CLASSIFY)
tool.merge_df_dataset(f_other, conf.HDF5_OTHER_CODE_CLASSIFY,
code_classify_df)
f_other.close()
return
def _get_merge_df(df, multi, ktype):
"""
根据获取的数据，进行聚合(需要注意聚合数据的起点)
"""
merge_df = tool.init_empty_df(None)
for index, row in df.iterrows():
datetime_obj = datetime.strptime(row[conf.HDF5_SHARE_DATE_INDEX],
"%Y-%m-%d %H:%M:%S")
if (ktype == '30' and datetime_obj.minute % 30
== 0) or (ktype == 'H' and datetime_obj.hour % 4 == 0):
if (index + multi - 1) < len(df):
row_dict = dict()
row_dict['date'] = row['date']
row_dict['open'] = row['open']
row_dict['close'] = df.iloc[index + multi - 1]['close']
row_dict['high'] = row['high']
row_dict['low'] = row['low']
row_dict['volume'] = row['volume']
for i in range(1, multi):
one = df.iloc[index + i]
row_dict['volume'] += one['volume']
row_dict['high'] = max(one['high'], row_dict['high'])
row_dict['low'] = min(one['low'], row_dict['low'])
merge_df = merge_df.append(row_dict, ignore_index=True)
else:
continue
return merge_df
class Wrap(object):
# 存储数据集合
raw_df = None
wrap_df = None
df = tool.init_empty_df([INDEX_DATE, INDEX_VALUE])
def __init__(self, df):
self.raw_df = df
self.wrap_df = tool.init_empty_df(self.raw_df.columns)
return
def merge_line(self, high_column, low_column, merge_price):
for index, row in self.raw_df.iterrows():
if index == 0:
self.wrap_df = self.wrap_df.append(row)
continue
length = len(self.wrap_df)
pre_row = self.wrap_df.iloc[-1]
# 递归向前合并?
if (pre_row[high_column] <= row[high_column] and pre_row[low_column] >= row[low_column]) or \
(pre_row[high_column] >= row[high_column] and pre_row[low_column] <= row[low_column]):
# 上升过程中，N-1被N包含，两K线最高点当高点，低点中较高者当低点
# 下降过程中，N-1包含N，两K线最低点当低点，高点中较低者当高点
self.wrap_df.loc[length - 1:length, high_column] = row[high_column]
# TODO 合并后，向前检查直到无法合并
for i in range(1, (len(self.wrap_df) - 1)):
for_pre_row = self.wrap_df.iloc[-i - 1]
for_row = self.wrap_df.iloc[-i]
if (for_pre_row[high_column] <= for_row[high_column] and for_pre_row[low_column] >= for_row[low_column]) or \
(for_pre_row[high_column] >= for_row[high_column] and for_pre_row[low_column] <= for_row[low_column]):
self.wrap_df.loc[length - 1 - i:length - i, high_column] = for_row[high_column]
else:
self.wrap_df = self.wrap_df.head(len(self.wrap_df) - i)
break
else:
self.wrap_df = self.wrap_df.append(row)
self.wrap_df = self.wrap_df.reset_index(drop=True)
if merge_price is True:
return self._merge_price(high_column, low_column)
else:
return self.wrap_df
def _merge_price(self, high_column, low_column):
# 将open、close、high、low聚合成一个数值
for index, row in self.wrap_df.iterrows():
one = dict()
one[INDEX_DATE] = row[INDEX_DATE]
if index == 0:
one[INDEX_VALUE] = row[high_column]
continue
pre_row = self.wrap_df.iloc[-1]
if pre_row[low_column] < row[low_column] and pre_row[high_column] < row[high_column]:
# 如果上涨，取high作为value
one[INDEX_VALUE] = row[high_column]
else:
# 如果下跌，取low作为value
one[INDEX_VALUE] = row[low_column]
self.df = self.df.append(one, ignore_index=True)
return self.df
def _get_data_by_spider(symbol, ktype, start, end):
"""
根据爬虫地址获取
"""
client = bitmexClient.Client(conf.BITMEX_URL_HISTORY)
params = {
"symbol": symbol,
"resolution": ktype,
"from": start,
"to": end,
}
data_json = client.get(params)
df = tool.init_empty_df(SYMBOL_COLS)
if data_json['s'] == 'ok':
open_arr = data_json['o']
close_arr = data_json['c']
high_arr = data_json['h']
low_arr = data_json['l']
time_arr = data_json['t']
volume_arr = data_json['v']
for index in range(0, len(time_arr)):
row_dict = dict()
row_dict['date'] = tradetime.transfer_unixtime(
time_arr[index], ktype)
row_dict['open'] = float(open_arr[index])
row_dict['close'] = float(close_arr[index])
row_dict['high'] = float(high_arr[index])
row_dict['low'] = float(low_arr[index])
row_dict['volume'] = volume_arr[index]
df = df.append(row_dict, ignore_index=True)
if len(df) > 0:
df["volume"] = df["volume"].astype('float64')
return df
def merge(self, ptype):
if ptype == PTYPE_MACD:
self.phase_df = tool.init_empty_df(MACD_COLUMNS)
ret = self.merge_macd("close")
else:
ret = None
return ret
def list(symbol, count, start, filter_dict=None):
"""
交易历史列表
"""
client = bitmexClient.Client(conf.BITMEX_URL_ORDER)
params = {
"symbol": symbol,
"count": count,
"start": start,
"reverse": True
}
if filter_dict is not None:
params['filter'] = filter_dict
data_json = client.get(params)
df = tool.init_empty_df(None)
for one in data_json:
row_dict = dict()
row_dict['date'] = tradetime.transfer_iso_datetime(
one['timestamp'], "M")
row_dict['symbol'] = one['symbol']
row_dict['side'] = one['side']
row_dict['type'] = one['ordType']
row_dict['status'] = one['ordStatus']
row_dict['price'] = one['price']
row_dict['order'] = one['simpleOrderQty']
row_dict['cum'] = one['simpleCumQty']
row_dict['id'] = one['orderID']
df = df.append(row_dict, ignore_index=True)
return df
def central(share_df):
"""
聚合k线后，标记顶与底，统计震荡中枢数量
"""
wrap_df = trans_wrap(share_df, False)
phase_df = point_phase(wrap_df)
# 如果聚合区间不足三条，无法进行分析
if len(phase_df) <= 3:
return None
# 根据聚合的up与down的值，判断是否存在中枢
central_dict = dict()
central_df = tool.init_empty_df(
["high", "low", "count", "start_date", "end_date"])
for index, row in phase_df.iterrows():
if index == 0:
central_dict["high"] = row["high"]
central_dict["low"] = row["low"]
central_dict["count"] = 1
central_dict["out_count"] = 0
central_dict["start_date"] = row["start_date"]
else:
compare_high = min(row["high"], central_dict["high"])
compare_low = max(row["low"], central_dict["low"])
if compare_high < compare_low:
if central_dict["count"] < 3:
# 不存在重叠，重置数据
pre_row = phase_df.iloc[index - 1]
central_dict["count"] = 1
central_dict["out_count"] = 0
central_dict["high"] = min(row["high"], pre_row["high"])
central_dict["low"] = max(row["low"], pre_row["low"])
central_dict["start_date"] = row["start_date"]
else:
# 已重叠，并出现脱离中枢范围的情况
central_dict["out_count"] += 1
if central_dict["out_count"] >= 2:
central_dict["end_date"] = phase_df.iloc[index -
2]["end_date"]
central_df = central_df.append(central_dict,
ignore_index=True)
pre_row = phase_df.iloc[index - 1]
central_dict["high"] = min(row["high"],
pre_row["high"])
central_dict["low"] = max(row["low"], pre_row["low"])
central_dict["start_date"] = pre_row["start_date"]
central_dict["count"] = 2
central_dict["out_count"] = 0
else:
# 存在重叠
central_dict["high"] = compare_high
central_dict["low"] = compare_low
central_dict["count"] += 1
# 前三根，判断是否存在重叠范围，如果存在则出现中枢
# 连续两根超出中枢范围，进入新的趋势判断
central_dict = central_dict.drop("out_count", axis=1)
return central_dict
def book(symbol, depth):
client = bitmexClient.Client(conf.BITMEX_URL_ORDERBOOK)
params = {"symbol": symbol, "depth": depth}
data_json = client.get(params)
df = tool.init_empty_df(None)
for one in data_json:
row_dict = dict()
row_dict['price'] = one['price']
row_dict['side'] = one['side']
row_dict['size'] = one['size']
df = df.append(row_dict, ignore_index=True)
return df
def code_detail(code_list, start_date):
"""
将code的basic内容，整理至share文件下
"""
# 获取basic所有日期的detail，并遍历读取详细信息
f = h5py.File(conf.HDF5_FILE_BASIC, 'a')
f_share = h5py.File(conf.HDF5_FILE_SHARE, 'a')
console.write_head(conf.HDF5_OPERATE_ARRANGE, conf.HDF5_RESOURCE_TUSHARE,
conf.HDF5_BASIC_DETAIL)
path = '/' + conf.HDF5_BASIC_DETAIL
if f.get(path) is None:
return
code_basic_dict = dict()
for date in f[path]:
if start_date is not None and date < start_date:
console.write_msg(start_date + "起始日期大于基本数据的最大日期")
continue
df = tool.df_from_dataset(f[path], date, None)
df["code"] = df["code"].str.decode("utf-8")
df = df.set_index("code")
for code in df.index:
if code not in code_list:
continue
if code not in code_basic_dict:
code_basic_dict[code] = tool.init_empty_df(df.columns)
code_basic_dict[code].loc[date] = df.loc[code, :]
for code, code_df in code_basic_dict.items():
code_df.index.name = conf.HDF5_SHARE_DATE_INDEX
code_df = code_df.reset_index().sort_values(
by=[conf.HDF5_SHARE_DATE_INDEX])
code_prefix = code[0:3]
code_group_path = '/' + code_prefix + '/' + code
if f_share.get(code_group_path) is None:
console.write_msg(code + "的detail文件不存在")
continue
if start_date is None:
tool.delete_dataset(f_share[code_group_path],
conf.HDF5_BASIC_DETAIL)
tool.merge_df_dataset(f_share[code_group_path], conf.HDF5_BASIC_DETAIL,
code_df)
console.write_exec()
console.write_blank()
console.write_tail()
f_share.close()
f.close()
return
def one_classify_detail(f, code_list, omit_list, ktype, start_date):
"""
根据单个分类，聚合所有code获取分类均值
"""
# 初始化一个全日期的空DataFrame，并初始化一列作为统计个数
init_df = tool.init_empty_df(conf.HDF5_SHARE_COLUMN)
# 按照列表顺序，获取code并逐一添加至初始化DF，并递增该日期的个数
for row in code_list:
code = row[0].astype(str)
code_prefix = code[0:3]
# 判断是否跳过创业板
if code_prefix in omit_list:
continue
code_group_path = '/' + code_prefix + '/' + code
if f.get(code_group_path) is None:
console.write_msg(code + "的detail文件不存在")
continue
else:
# 忽略停牌、退市、无法获取的情况
if f[code_group_path].attrs.get(conf.HDF5_BASIC_QUIT) is not None:
continue
if f[code_group_path].attrs.get(conf.HDF5_BASIC_ST) is not None:
continue
if f[code_group_path].get(ktype) is None:
console.write_msg(code + "的" + ktype + "文件不存在")
continue
add_df = tool.df_from_dataset(f[code_group_path], ktype, None)
add_df[conf.HDF5_SHARE_DATE_INDEX] = add_df[
conf.HDF5_SHARE_DATE_INDEX].str.decode("utf-8")
add_df["num"] = 1
add_df = add_df.set_index(conf.HDF5_SHARE_DATE_INDEX)
init_df = init_df.add(add_df, fill_value=0)
# 总数除以数量，得到平均值
if len(init_df) > 0:
init_df = init_df.div(init_df.num, axis=0)
init_df = init_df.drop("num", axis=1)
if start_date is not None:
init_df = init_df.ix[start_date:]
init_df = init_df.reset_index().sort_values(
by=[conf.HDF5_SHARE_DATE_INDEX])
init_df["volume"] = init_df["volume"].astype('float64')
return init_df
else:
return None
def point_phase(wrap_df):
point_df = point(wrap_df[["high", "low", "date"]])
# 取up的high，down的low值，组成一个阶段的up与down值，逐个阶段判断是否存在重合
phase_df = tool.init_empty_df(["high", "low", "start_date", "end_date"])
row_dict = dict()
for index, row in point_df.iterrows():
# 如果第一个尖点是底则忽略
if index == 0 and row[INDEX_POINT_TYPE] == POINT_UP:
continue
if row[INDEX_POINT_TYPE] == POINT_DOWN:
row_dict["low"] = row["low"]
row_dict["start_date"] = row["date"]
else:
row_dict["high"] = row["high"]
row_dict["end_date"] = row["date"]
phase_df = phase_df.append(row_dict, ignore_index=True)
row_dict = dict()
return phase_df
def wallet_history(count, start):
"""
钱包历史列表
"""
client = bitmexClient.Client(conf.BITMEX_URL_WALLET_HISTORY)
params = {"count": count, "start": start, "reverse": True}
data_json = client.get(params)
df = tool.init_empty_df(WALLET_COLS)
for one in data_json:
row_dict = dict()
row_dict['status'] = one['transactStatus']
row_dict['address'] = one['address']
row_dict['amount'] = one['amount']
row_dict['fee'] = one['fee']
row_dict['balance'] = one['walletBalance']
row_dict['date'] = tradetime.transfer_iso_datetime(
one['timestamp'], "M")
df = df.append(row_dict, ignore_index=True)
return df
def _transfer_json_to_df(data_json):
"""
将返回的json转化为df
"""
df = tool.init_empty_df(None)
for one in data_json:
row_dict = dict()
row_dict['date'] = tradetime.transfer_iso_datetime(
one['timestamp'], "M")
row_dict['open'] = one['open']
row_dict['low'] = one['low']
row_dict['high'] = one['high']
row_dict['close'] = one['close']
row_dict['volume'] = one['volume']
# row_dict['turnover'] = one['turnover']
df = df.append(row_dict, ignore_index=True)
if len(df) > 0:
df["volume"] = df["volume"].astype('float64')
df = df.reindex(index=df.index[::-1]).reset_index(drop=True)
return df
class strategy(object):
"""
大-中-小，三重级别递归策略
1. 大与中决定多空方向，小决定交易节点
2. 小级别的phase背离出现次数，决定杠杆倍率，并推动中级别方向转换
3. 大或中出现trend背离，子级别方向转换压制
例如父级别趋势向下，1min出现买点，则起步选择做多
子级别第一个买点杠杆1x，上涨后macd转折出售，如果dea与dif未上传零轴，5min方向不变
第二个买点出现，正常买点杠杆不变，背离买点杠杆增加至2x，背离杠杆最多增加至3x
期间子级别的相反趋势，可以用0.5x杠杆进行交易
当子级别出现卖点背离时，父级别趋势逆转，方向由做多转为做空
"""
# 用于存储code名称
code = None
# 数据来源类别
stype = None
# 回测标签，回测时不更新数据源
backtest = None
# 是否将最新数据回写至文件
rewrite = None
# 父级别方向
medium_side = None
# macd波动因子
factor_macd_range = None
# 大级别对应时间节点
big_level = None
# 大级别trend背离标签
big_trend_reverse = False
# 存储大级别数据
big = None
# 中级别对应时间节点
medium_level = None
# 中级别trend背离标签
medium_trend_reverse = False
# 存储中级别数据
medium = None
# 小级对应时间节点
small_level = None
# 小级别phase背离标签
small_phase_reverse = False
# 存储小级别数据
small = None
# 存储小级别趋势数据
phase = tool.init_empty_df(PHASE_COLUMNS)
def __init__(self,
code,
stype,
backtest,
rewrite,
small_level,
factor_macd_range=0.1):
self.code = code
self.stype = stype
self.backtest = backtest
self.rewrite = rewrite
self.small_level = small_level
self.factor_macd_range = factor_macd_range
if stype == conf.STYPE_BITMEX:
self.medium_level = BITMEX_LEVEL_DICT[self.small_level]
self.big_level = BITMEX_LEVEL_DICT[self.medium_level]
return
def prepare(self):
"""
初始化数据
"""
update_dict = {
DF_SMALL: self.small_level,
DF_MEDIUM: self.medium_level,
DF_BIG: self.big_level,
}
num_dict = {
self.small_level: 180,
self.medium_level: 48,
self.big_level: 48,
}
for key in update_dict:
ktype = update_dict[key]
file_num = num_dict[ktype]
direct_turn = False
if ktype in [conf.BINSIZE_ONE_MINUTE]:
direct_turn = False
df = trend.get_from_file(ktype, conf.STYPE_BITMEX, self.code,
self.factor_macd_range, file_num,
direct_turn)
setattr(self, key, df)
return
def update(self):
update_dict = {
DF_SMALL: self.small_level,
DF_MEDIUM: self.medium_level,
DF_BIG: self.big_level,
}
for key in update_dict:
# 以df最后的时间为准
last_date = getattr(self, key).iloc[-1][conf.HDF5_SHARE_DATE_INDEX]
ktype = update_dict[key]
pull_flag = tradetime.check_pull_time(last_date, ktype)
if pull_flag is False:
continue
new_df = trend.get_from_remote(ktype, conf.STYPE_BITMEX, last_date,
self.code, self.rewrite)
# 更新macd趋势列表
trend_df = getattr(self, key)
df_length = len(trend_df)
direct_turn = False
if ktype in [conf.BINSIZE_ONE_MINUTE]:
direct_turn = False
trend_df = trend.append_and_macd(trend_df, new_df, last_date,
self.factor_macd_range,
direct_turn)
setattr(self, key, trend_df.tail(df_length).reset_index(drop=True))
return
def check_all(self):
"""
遍历子级别macd趋势，并罗列买卖信号
"""
trend_df = self.small
for i in range(3, len(trend_df) + 1):
self.small = trend_df.head(i)
result = self.check_new()
if result is not False:
self.output()
self.small = trend_df
return
def check_new(self):
"""
检查最新子级别macd趋势
"""
# 更新趋势聚合
self.merge_phase()
now = self.small.iloc[-1]
now_date = now[conf.HDF5_SHARE_DATE_INDEX]
# 检查medium与big是否存在trend背离
check_dict = [DF_MEDIUM, DF_BIG]
for df_name in check_dict:
check_df = getattr(self, df_name)
result = trend.check_reverse(now_date, check_df)
setattr(self, df_name + "_" + TREND_REVERSE, result)
# TODO 检查small的phase背离
# 获取最新的phase状态
phase_now = self.phase.iloc[-1]
phase_status = phase_now[action.INDEX_PHASE_STATUS]
phase_date = phase_now[conf.HDF5_SHARE_DATE_INDEX]
ret = False
if phase_date == now_date:
ret = self._get_side(phase_status)
return ret
def merge_phase(self):
self.phase = phase.latest_dict(self.small, self.phase)
return
def output(self):
"""
打印结果
"""
trend_now = self.small.iloc[-1]
phase_now = self.phase.iloc[-1]
# phase_pre = self.phase.iloc[-2]
side = self._get_side(phase_now[action.INDEX_PHASE_STATUS])
console.write_msg(
"【%s, %s, %s】" %
(self.code, trend_now[conf.HDF5_SHARE_DATE_INDEX], side))
if len(self.phase) > 1:
phase_pre = self.phase.iloc[-2]
msg = "上阶段，macd差值%f，price差值%d，连续%d次，dif位置%f"
console.write_msg(
msg % (phase_pre[phase.MACD_DIFF], phase_pre[phase.PRICE_END] -
phase_pre[phase.PRICE_START], phase_pre[phase.COUNT],
phase_pre[phase.DIF_END]))
msg = "建议价格%d - %d"
console.write_msg(msg %
(phase_now[phase.PRICE_START], trend_now["close"]))
msg = "medium级别背离情况：%s"
console.write_msg(msg % (self.medium_trend_reverse))
msg = "big级别背离情况：%s"
console.write_msg(msg % (self.big_trend_reverse))
console.write_blank()
return
def _get_side(self, phase_status):
if phase_status == action.STATUS_UP:
ret = conf.BUY_SIDE
else:
ret = conf.SELL_SIDE
return ret
class strategy(object):
"""
5min级别的macd趋势震荡策略
核心:
1. 状态判断(震荡因子):5min的macd-shake幅度
2. 仓位控制(比例因子):大方向(例如a股日线,bitmex的4h线)、30min级别、指数的30min级别
3. 交易时机(未来因子):1min的背离
"""
# 用于存储code名称
code = None
# 数据来源类别
stype = None
# 回测标签，回测时不更新数据源
backtest = None
# 是否将最新数据回写至文件
rewrite = None
# 5min是否处于shake状态
five_shake = False
# 5min的macd趋势是否处于背离
five_trend_reverse = False
# 1min的macd阶段是否存在背离
one_phase_reverse = False
# 存储1m数据
one = None
# 存储5m数据
five = None
# 存储30m数据
thirty = None
# 存储大方向(例如a股日线、bitmex的4h线)数据
big = None
# 存储相关指数的30m数据
index = None
# 存储处理得到的买卖点
trade = tool.init_empty_df(None)
# 震荡因子，shake幅度百分比，默认10%
factor_macd_range = 0.1
def __init__(self, code, stype, backtest, rewrite, factor_macd_range=None):
self.code = code
self.stype = stype
self.backtest = backtest
self.rewrite = rewrite
if factor_macd_range is not None:
self.factor_macd_range = factor_macd_range
return
def prepare(self):
"""
初始化数据
"""
if self.stype == conf.STYPE_ASHARE:
ktype_dict = {
DF_FIVE: conf.KTYPE_FIVE,
DF_THIRTY: conf.KTYPE_THIRTY,
DF_BIG: conf.KTYPE_DAY,
DF_INDEX: conf.KTYPE_THIRTY,
}
code_prefix = self.code[0:1]
if code_prefix == "0":
index = "sz"
elif code_prefix == "6":
index = "sh"
elif self.stype == conf.STYPE_BITMEX:
ktype_dict = {
DF_FIVE: conf.BINSIZE_FIVE_MINUTE,
DF_THIRTY: conf.BINSIZE_THIRTY_MINUTE,
DF_BIG: conf.BINSIZE_FOUR_HOUR,
DF_INDEX: conf.BINSIZE_THIRTY_MINUTE,
}
index = conf.BITMEX_BXBT
else:
raise Exception("stype未配置或配置错误")
for key in ktype_dict:
ktype = ktype_dict[key]
if key == DF_INDEX:
setattr(
self, key,
trend.get_from_file(ktype, self.stype, index,
self.factor_macd_range))
else:
setattr(
self, key,
trend.get_from_file(ktype, self.stype, self.code,
self.factor_macd_range))
if self.backtest is False:
self.update()
return
def update(self):
if self.stype == conf.STYPE_ASHARE:
ktype_dict = {
DF_FIVE: conf.KTYPE_FIVE,
DF_THIRTY: conf.KTYPE_THIRTY,
DF_BIG: conf.KTYPE_DAY,
DF_INDEX: conf.KTYPE_THIRTY,
}
code_prefix = self.code[0:1]
if code_prefix == "0":
index = "sz"
elif code_prefix == "6":
index = "sh"
elif self.stype == conf.STYPE_BITMEX:
ktype_dict = {
DF_FIVE: conf.BINSIZE_FIVE_MINUTE,
DF_THIRTY: conf.BINSIZE_THIRTY_MINUTE,
DF_BIG: conf.BINSIZE_FOUR_HOUR,
DF_INDEX: conf.BINSIZE_THIRTY_MINUTE,
}
index = conf.BITMEX_BXBT
else:
raise Exception("stype未配置或配置错误")
for key in ktype_dict:
# 以df最后的时间为准
last_date = getattr(self, key).iloc[-1][conf.HDF5_SHARE_DATE_INDEX]
# 根据时间节点判断是否拉取
ktype = ktype_dict[key]
if self.rewrite is True:
pull_flag = tradetime.check_pull_time(last_date, ktype)
if pull_flag is False:
continue
if key == DF_INDEX:
new_df = trend.get_from_remote(ktype, self.stype, last_date,
index, self.rewrite)
else:
new_df = trend.get_from_remote(ktype, self.stype, last_date,
self.code, self.rewrite)
# 定期更新trend_df
if ktype != conf.KTYPE_DAY and self.rewrite is False:
new_df[conf.HDF5_SHARE_DATE_INDEX] = new_df[
conf.HDF5_SHARE_DATE_INDEX].apply(
lambda x: tradetime.transfer_date(x, ktype, "S"))
trend_df = getattr(self, key)
trend_df = trend.append_and_macd(trend_df, new_df, last_date,
self.factor_macd_range)
setattr(self, key, trend_df.reset_index(drop=True))
return
def check_all(self):
"""
遍历5min-macd所有趋势，罗列历史买卖信号
"""
trend_df = self.five
for i in range(3, len(trend_df) + 1):
tmp_df = trend_df.head(i)
self.five = tmp_df
result = self.check_new()
if result is True:
self.save_trade()
self.output()
self.five = trend_df
return
def check_new(self):
"""
检查5min-macd最新趋势，是否存在买卖信号
"""
now = self.five.iloc[-1]
pre = self.five.iloc[-2]
phase_start, phase_end = phase.now(self.five)
if phase_start is None:
return False
# 当前macd趋势的macd波动幅度
phase_range = phase_end["macd"] - phase_start["macd"]
# 当前macd趋势的price波动幅度
price_range = phase_end["close"] - phase_start["close"]
# 背离分析，如果macd是下降趋势，但是价格上涨，则属于背离；反之同理
self.five_trend_reverse = False
if (phase_range > 0 and price_range < 0) or (phase_range < 0
and price_range > 0):
self.five_trend_reverse = True
# check检查的波动，在计算基础上扩张一部分
macd_range = self.factor_macd_range * abs(phase_range) * 1
ret = False
if self.five_shake is False:
# 判断第一次出现的波动，如果该段趋势太小则没有做T必要
# TODO phase_range的大小判断，避免判断太小的range趋势(目前用的是旧方案，trend数量)
if phase_end[action.INDEX_TREND_COUNT] >= 5 and now[
action.INDEX_STATUS] == action.STATUS_SHAKE and pre[
action.INDEX_STATUS] != action.STATUS_SHAKE:
# 检查macd波动幅度
macd_diff = abs(now["macd"] - pre["macd"])
if macd_diff > macd_range:
# 如果macd波动值超出范围，视为转折
ret = True
else:
# 如果macd波动值在范围以内，则视为波动，观察后续走势
self.five_shake = True
else:
# 判断波动过程
if now["status"] != action.STATUS_SHAKE:
# 波动结束趋势逆转
self.five_shake = False
# 跟shake前的状态进行比较
for i in range(2, now[action.INDEX_TREND_COUNT] + 2):
shake_before = self.five.iloc[-i]
if shake_before[
action.INDEX_STATUS] == action.STATUS_SHAKE:
continue
else:
break
if now["status"] != shake_before[action.INDEX_STATUS]:
ret = True
else:
macd_diff = abs(now["macd"] - phase_end["macd"])
# 波动超出边界，震荡结束
if macd_diff > macd_range:
self.five_shake = False
ret = True
return ret
def reverse(self):
"""
定位1min-phase背离
"""
return
def save_trade(self):
"""
存储买卖点
"""
# 获取当前5min非shake的状态
trade_dict = dict()
now = self.five.iloc[-1]
phase_start, phase_end = phase.now(self.five)
phase_range = abs(phase_end["macd"] - phase_start["macd"])
if now[action.INDEX_STATUS] == action.STATUS_SHAKE:
pre_phase_status = phase_end[action.INDEX_STATUS]
else:
pre_phase_status = phase_start[action.INDEX_STATUS]
macd_diff = abs(now["macd"] - phase_end["macd"])
# 判断交易点性质
if pre_phase_status == action.STATUS_UP:
if self.five_trend_reverse is True:
trade_type = "背离买点"
else:
trade_type = "正常卖点"
else:
if self.five_trend_reverse is True:
trade_type = "背离卖点"
else:
trade_type = "正常买点"
trade_dict = dict()
trade_dict[conf.HDF5_SHARE_DATE_INDEX] = now[
conf.HDF5_SHARE_DATE_INDEX]
trade_dict[TRADE_TYPE] = trade_type
trade_dict[DF_FIVE + TRADE_STATUS] = pre_phase_status
trade_dict[DF_FIVE + TRADE_MACD_PHASE] = phase_range
trade_dict[DF_FIVE + TRADE_MACD_DIFF] = round(
macd_diff * 100 / phase_range, 0)
trade_dict[DF_FIVE +
TRADE_TREND_COUNT] = phase_end[action.INDEX_TREND_COUNT]
# 获取仓位与状态
positions = 0
dtype_dict = {
DF_THIRTY: "个股30min",
DF_INDEX: "指数30min",
DF_BIG: "个股大趋势",
}
for dtype in dtype_dict:
status, trend_count, macd_diff, phase_range = self.get_relate_status(
dtype)
positions = self.count_positions(pre_phase_status, status,
positions)
trade_dict[dtype + TRADE_STATUS] = status
trade_dict[dtype + TRADE_MACD_PHASE] = phase_range
trade_dict[dtype + TRADE_MACD_DIFF] = round(
macd_diff * 100 / phase_range, 0)
trade_dict[dtype + TRADE_TREND_COUNT] = trend_count
trade_dict[TRADE_POSITIONS] = positions
# 更新trade列表
self.trade = self.trade.append(trade_dict, ignore_index=True)
return
def output(self):
now = self.trade.iloc[-1]
console.write_msg(
"【%s, %s, %s】" %
(self.code, now[conf.HDF5_SHARE_DATE_INDEX], now["type"]))
msg = "个股5min，趋势%s，连续%d次，macd趋势%f, macd差值%d%%"
console.write_msg(
msg %
(now[DF_FIVE + TRADE_STATUS], now[DF_FIVE + TRADE_TREND_COUNT],
now[DF_FIVE + TRADE_MACD_PHASE], now[DF_FIVE + TRADE_MACD_DIFF]))
dtype_dict = {
DF_THIRTY: "个股30min",
DF_BIG: "个股大趋势",
DF_INDEX: "指数30min",
}
for dtype in dtype_dict:
msg = "%s，趋势%s，连续%d次，macd趋势%f, macd差值%d%%"
console.write_msg(
msg %
(dtype_dict[dtype], now[dtype + TRADE_STATUS],
now[dtype + TRADE_TREND_COUNT], now[dtype + TRADE_MACD_PHASE],
now[dtype + TRADE_MACD_DIFF]))
# TODO 背离情况要计算两次买卖点trend_count的差值
trend_count = now[DF_FIVE + TRADE_TREND_COUNT]
upper_estimate = (trend_count + 1) * 5
lower_estimate = (trend_count - 1) * 5
if self.stype == conf.STYPE_ASHARE:
remain_seconds = tradetime.get_ashare_remain_second(
now[conf.HDF5_SHARE_DATE_INDEX])
remain_minutes = round(remain_seconds / 60, 0)
msg = "剩余时间%d分钟，下个交易点预估需要%d-%d分钟，模式%s"
if (upper_estimate * 60) <= remain_seconds:
trade_opportunity = "T+0"
else:
trade_opportunity = "T+1"
console.write_msg(msg % (remain_minutes, lower_estimate,
upper_estimate, trade_opportunity))
elif self.stype == conf.STYPE_BITMEX:
msg = "下个交易点预估需要%d-%d分钟"
console.write_msg(msg % (lower_estimate, upper_estimate))
console.write_msg("建议仓位：%d/3" % (now[TRADE_POSITIONS]))
console.write_blank()
return
def count_positions(self, pre_status, status, positions):
"""
仓位估算
考虑相关趋势的影响，相反方向会产生压制，例如relate上升对5min卖点，relate下降对5min买点
"""
# 检查big 30min index的趋势,计算仓位
if pre_status == action.STATUS_UP:
if status == action.STATUS_UP:
positions += 0
elif status == action.STATUS_DOWN:
positions += 1
else:
status_arr = status.split("-")
if status_arr[0] == action.STATUS_UP:
positions += 1
else:
if status == action.STATUS_DOWN:
positions += 0
elif status == action.STATUS_UP:
positions += 1
else:
status_arr = status.split("-")
if status_arr[0] == action.STATUS_DOWN:
positions += 1
return positions
def get_relate_status(self, dtype):
"""
获取index, big, 30min的最新状态
"""
now_date = self.five.iloc[-1][conf.HDF5_SHARE_DATE_INDEX]
if dtype == DF_BIG:
trend_df = self.big
elif dtype == DF_THIRTY:
trend_df = self.thirty
elif dtype == DF_INDEX:
trend_df = self.index
trend_df = trend_df[trend_df[conf.HDF5_SHARE_DATE_INDEX] <= now_date]
now = trend_df.iloc[-1]
pre = trend_df.iloc[-2]
phase_start, phase_end = phase.now(trend_df)
# 如果最新状态是震荡，则追溯至非震荡状态的时间点，非震荡则直接获取状态
if now[action.INDEX_STATUS] == action.STATUS_SHAKE:
status = phase_end[action.INDEX_STATUS] + "-" + now[
action.INDEX_STATUS]
macd_diff = abs(now["macd"] - phase_end["macd"])
phase_range = abs(phase_end["macd"] - phase_start["macd"])
trend_count = phase_end[action.INDEX_TREND_COUNT]
else:
macd_diff = abs(now["macd"] - pre["macd"])
phase_range = abs(phase_end["macd"] - phase_start["macd"])
status = now[action.INDEX_STATUS]
trend_count = now[action.INDEX_TREND_COUNT]
return status, trend_count, macd_diff, phase_range
class Action(object):
# 存储index_df数据
index_df = None
# 存储所有时间节点的走向情况
df = tool.init_empty_df([
INDEX_DATE, INDEX_VALUE, INDEX_DIRECTION, INDEX_TURN_COUNT,
INDEX_TREND_COUNT, INDEX_ACTION, INDEX_STATUS, INDEX_PHASE_STATUS
])
# 转折后的上边界
up_border = 0
# 转折后的下边界
down_border = 0
# 方案A，绝对值(废弃):不同价格量级的数据(例如比特币)，波动幅度不同，不能用固定值
# 方案B，该段趋势波动差，差值越大越快说明趋势越急，对转折要求更高
factor_macd_range = None
# 直接转折标签，常用于1min等时间敏感节点
direct_turn = None
def __init__(self, index_df, factor_macd_range, direct_turn=False):
self.index_df = index_df
self.direct_turn = direct_turn
if factor_macd_range is not None:
self.factor_macd_range = factor_macd_range
else:
self.factor_macd_range = DEFAULT_FACTOR_MACD_RANGE
return
def all(self, date_column, value_column):
"""
获取从头到尾的趋势
"""
first = self.index_df.iloc[1]
second = self.index_df.iloc[2]
diff = second[value_column] - first[value_column]
if diff > 0 or (diff == 0 and second >= 0):
direction = DIRECTION_UP
status = STATUS_UP
elif diff < 0 or (diff == 0 and second < 0):
direction = DIRECTION_DOWN
status = STATUS_DOWN
first_row = {
INDEX_DATE: first[date_column],
INDEX_VALUE: first[value_column],
INDEX_DIRECTION: direction,
INDEX_STATUS: status,
INDEX_PHASE_STATUS: status,
INDEX_TURN_COUNT: 0,
INDEX_TREND_COUNT: 1,
INDEX_ACTION: TREND_STILL,
}
second_row = {
INDEX_DATE: second[date_column],
INDEX_VALUE: second[value_column],
INDEX_DIRECTION: direction,
INDEX_STATUS: status,
INDEX_PHASE_STATUS: status,
INDEX_TURN_COUNT: 0,
INDEX_TREND_COUNT: 2,
INDEX_ACTION: TREND_STILL
}
self.df = self.df.append(first_row, ignore_index=True)
self.df = self.df.append(second_row, ignore_index=True)
for index, row in self.index_df[3:].iterrows():
date = row[date_column]
value = row[value_column]
pre_turn_count = self.df.iloc[-1][INDEX_TURN_COUNT]
# 1.单边趋势
if pre_turn_count == 0:
one = self.trend(value, date)
# 2.第N次转折后，根据选择方向进行判断:
elif pre_turn_count >= 1:
one = self.turn(value, date)
self.df = self.df.append(one, ignore_index=True)
return self.df
def trend(self, value, date):
"""
单边趋势的方向选择
"""
one = dict()
one[INDEX_DATE] = date
one[INDEX_VALUE] = value
pre_row = self.df.iloc[-1]
# 趋势延续
if self.compare_border(TREND_STILL,
value) or (value == pre_row[INDEX_VALUE]):
self.reset_border()
one[INDEX_ACTION] = TREND_STILL
one[INDEX_TREND_COUNT] = pre_row[INDEX_TREND_COUNT] + 1
one[INDEX_TURN_COUNT] = 0
one[INDEX_STATUS] = pre_row[INDEX_STATUS]
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
one[INDEX_DIRECTION] = pre_row[INDEX_DIRECTION]
# 出现转折
elif self.compare_border(TREND_TURN, value):
pre_value = pre_row[INDEX_VALUE]
# 对于第一个趋势的bar，如果无法追溯最初的则标记
if len(self.df) > pre_row[INDEX_TREND_COUNT]:
start_row = self.df.iloc[-1 - pre_row[INDEX_TREND_COUNT]]
real_start = True
else:
start_row = self.df.iloc[-pre_row[INDEX_TREND_COUNT] + 1]
real_start = False
macd_range = abs(
(pre_value - start_row[INDEX_VALUE]) * self.factor_macd_range)
if self.direct_turn is True and abs(
pre_value -
value) > macd_range and real_start is True and pre_row[
INDEX_TREND_COUNT] > 4:
one[INDEX_ACTION] = TREND_STILL
one[INDEX_TREND_COUNT] = 1
one[INDEX_TURN_COUNT] = 0
one[INDEX_DIRECTION] = self.get_op_d(pre_row[INDEX_DIRECTION])
one[INDEX_STATUS] = self.get_d_s(one[INDEX_DIRECTION])
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
else:
if pre_row[INDEX_DIRECTION] == DIRECTION_UP:
self.set_border(pre_value - macd_range, pre_value)
else:
self.set_border(pre_value + macd_range, pre_value)
one[INDEX_ACTION] = TREND_TURN
one[INDEX_TREND_COUNT] = 1
one[INDEX_TURN_COUNT] = 1
one[INDEX_STATUS] = STATUS_SHAKE
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
one[INDEX_DIRECTION] = self.get_op_d(pre_row[INDEX_DIRECTION])
return one
def turn(self, value, date):
"""
翻转后的方向选择
"""
one = dict()
one[INDEX_DATE] = date
one[INDEX_VALUE] = value
pre_row = self.df.iloc[-1]
border_row = None
for i in range(1, len(self.df)):
border_row = self.df.iloc[-i]
if border_row[INDEX_STATUS] != STATUS_SHAKE:
break
# 如果未出现转折，并且超出转折边界
if self.compare_border(STILL_OUT, value):
one[INDEX_ACTION] = STILL_OUT
one[INDEX_DIRECTION] = pre_row[INDEX_DIRECTION]
one[INDEX_STATUS] = self.get_d_s(one[INDEX_DIRECTION])
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
one[INDEX_TREND_COUNT] = 1 + pre_row[INDEX_TREND_COUNT]
one[INDEX_TURN_COUNT] = 0
pre_turn_count = pre_row[INDEX_TURN_COUNT]
# 如果只转折一次
if pre_turn_count == 1:
if one[INDEX_TREND_COUNT] >= TURN_MIN_NUM or self.direct_turn is True:
self.df.loc[len(self.df) -
pre_row[INDEX_TREND_COUNT]:len(self.df),
INDEX_PHASE_STATUS] = self.get_d_s(
one[INDEX_DIRECTION])
self.reset_border()
else:
one[INDEX_STATUS] = STATUS_SHAKE
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
one[INDEX_TURN_COUNT] = 1
elif pre_turn_count % 2 == 0:
if pre_turn_count < SHAKE_MIN_NUM:
# 视作上一阶段延续
for i in range(1, len(self.df)):
if self.df.iloc[len(self.df) -
i][INDEX_PHASE_STATUS] != STATUS_SHAKE:
self.df.loc[len(self.df) - i:len(self.df),
INDEX_PHASE_STATUS] = self.get_d_s(
one[INDEX_DIRECTION])
break
one[INDEX_TREND_COUNT] = border_row[INDEX_TREND_COUNT] + i
self.reset_border()
else:
# 视作新的相反阶段开始
self.reset_border()
elif pre_turn_count % 2 == 1:
# 小于shake次数，趋势逆转，大于shake次数，震荡结束
if pre_turn_count < SHAKE_MIN_NUM:
self.reset_border()
else:
self.reset_border()
# 如果未出现转折，延伸长度未超出转折边界
elif self.compare_border(STILL_IN, value):
one[INDEX_ACTION] = STILL_IN
one[INDEX_TREND_COUNT] = pre_row[INDEX_TREND_COUNT] + 1
one[INDEX_TURN_COUNT] = pre_row[INDEX_TURN_COUNT]
one[INDEX_DIRECTION] = pre_row[INDEX_DIRECTION]
one[INDEX_STATUS] = pre_row[INDEX_STATUS]
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
# 如果出现转折，延伸长度未超出上一阶段极值
elif self.compare_border(TURN_IN, value):
one[INDEX_ACTION] = TURN_IN
one[INDEX_TREND_COUNT] = 1
one[INDEX_TURN_COUNT] = pre_row[INDEX_TURN_COUNT] + 1
one[INDEX_DIRECTION] = self.get_op_d(pre_row[INDEX_DIRECTION])
one[INDEX_STATUS] = STATUS_SHAKE
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
# 如果出现转折，超出上一阶段极值
elif self.compare_border(TURN_OUT, value):
one[INDEX_ACTION] = TURN_OUT
one[INDEX_TURN_COUNT] = 0
one[INDEX_DIRECTION] = self.get_op_d(pre_row[INDEX_DIRECTION])
one[INDEX_STATUS] = self.get_d_s(one[INDEX_DIRECTION])
one[INDEX_PHASE_STATUS] = one[INDEX_STATUS]
one[INDEX_TREND_COUNT] = 1
pre_turn_count = pre_row[INDEX_TURN_COUNT]
if pre_turn_count < SHAKE_MIN_NUM:
if pre_turn_count % 2 == 1:
# 上一阶段延续
self.df.loc[len(self.df) -
pre_row[INDEX_TREND_COUNT]:len(self.df),
INDEX_PHASE_STATUS] = self.get_d_s(
one[INDEX_DIRECTION])
one[INDEX_TREND_COUNT] = border_row[INDEX_TREND_COUNT] + i
else:
# 趋势逆转
for i in range(1, 100):
if self.df.iloc[len(self.df) -
i][INDEX_PHASE_STATUS] != STATUS_SHAKE:
self.df.loc[len(self.df) - i:len(self.df),
INDEX_PHASE_STATUS] = self.get_d_s(
one[INDEX_DIRECTION])
break
self.reset_border()
return one
def compare_border(self, atype, value):
"""
比较value与边界
"""
pre_row = self.df.iloc[-1]
d = pre_row[INDEX_DIRECTION]
pre = pre_row[INDEX_VALUE]
up_border = self.up_border
down_border = self.down_border
up_switch = {
TREND_STILL: value >= pre,
TREND_TURN: value < pre,
STILL_OUT: value > up_border,
STILL_IN: pre <= value <= up_border,
TURN_IN: down_border <= value < pre,
TURN_OUT: value < down_border,
}
down_switch = {
TREND_STILL: value <= pre,
TREND_TURN: value > pre,
STILL_OUT: value < down_border,
STILL_IN: pre >= value >= down_border,
TURN_IN: up_border >= value > pre,
TURN_OUT: value > up_border,
}
if d == DIRECTION_UP:
return up_switch[atype]
elif d == DIRECTION_DOWN:
return down_switch[atype]
else:
raise Exception("边界获取异常")
def set_border(self, a, b):
"""
设置初始边界
"""
self.up_border = max(a, b)
self.down_border = min(a, b)
return
def reset_border(self):
self.up_border = 0
self.down_border = 0
return
def get_op_d(self, d):
"""
获取相反方向
"""
if d == DIRECTION_UP:
return DIRECTION_DOWN
elif d == DIRECTION_DOWN:
return DIRECTION_UP
def get_d_s(self, d):
"""
获取方向的对应状态
"""
if d == DIRECTION_UP:
return STATUS_UP
elif d == DIRECTION_DOWN:
return STATUS_DOWN
def __init__(self, df):
self.raw_df = df
self.wrap_df = tool.init_empty_df(self.raw_df.columns)
return
def mark_grade(today_str=None):
"""
对筛选结果进行打分
"""
console.write_head(conf.HDF5_OPERATE_SCREEN, conf.HDF5_RESOURCE_TUSHARE,
conf.SCREEN_SHARE_GRADE)
f = h5py.File(conf.HDF5_FILE_SCREEN, 'a')
f_share = h5py.File(conf.HDF5_FILE_SHARE, 'a')
if today_str is None:
today_str = tradetime.get_today()
if f[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_FILTER].get(
today_str) is None:
console.write_msg(today_str + "个股筛选结果不存在")
return
grade_df = tool.init_empty_df([
"code", "status", "d_price_space", "d_price_per", "30_price_space",
"30_price_per", "d_macd", "30_macd"
])
screen_df = tool.df_from_dataset(
f[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_FILTER],
today_str, None)
screen_df["d_m_status"] = screen_df["d_m_status"].str.decode("utf-8")
screen_df["w_m_status"] = screen_df["w_m_status"].str.decode("utf-8")
screen_df["m_m_status"] = screen_df["m_m_status"].str.decode("utf-8")
screen_df["code"] = screen_df["code"].str.decode("utf-8")
for index, row in screen_df.iterrows():
code = row["code"]
grade_dict = dict()
grade_dict["code"] = code
grade_dict["status"] = 0
grade_dict["status"] += _status_grade(row["d_m_status"])
grade_dict["status"] += _status_grade(row["w_m_status"])
grade_dict["status"] += _status_grade(row["m_m_status"])
code_prefix = code[0:3]
code_group_path = '/' + code_prefix + '/' + code
for ktype in ["D", "30"]:
detail_ds_name = ktype
index_ds_name = conf.HDF5_INDEX_DETAIL + "_" + ktype
if f_share[code_group_path].get(detail_ds_name) is None:
console.write_msg(code + "的detail数据不存在")
continue
if f_share[code_group_path].get(index_ds_name) is None:
console.write_msg(code + "的index数据不存在")
continue
detail_df = tool.df_from_dataset(f_share[code_group_path],
detail_ds_name, None)
index_df = tool.df_from_dataset(f_share[code_group_path],
index_ds_name, None)
latest_price = detail_df.tail(1)["close"].values[0]
latest_macd = index_df.tail(1)["macd"].values[0]
diverse_price_start = row[str.lower(ktype) +
INDEX_MACD_DIVERSE_PRICE_START]
if diverse_price_start == 0:
grade_dict[str.lower(ktype) + "_price_space"] = 0
grade_dict[str.lower(ktype) + "_price_per"] = 0
else:
grade_dict[str.lower(ktype) + "_price_space"] = round(
diverse_price_start - latest_price, 2)
grade_dict[str.lower(ktype) + "_price_per"] = round(
grade_dict[str.lower(ktype) + "_price_space"] * 100 /
diverse_price_start, 2)
grade_dict[str.lower(ktype) + "_macd"] = latest_macd
grade_df = grade_df.append(grade_dict, ignore_index=True)
if f[conf.STRATEGY_TREND_AND_REVERSE].get(conf.SCREEN_SHARE_GRADE) is None:
f[conf.STRATEGY_TREND_AND_REVERSE].create_group(
conf.SCREEN_SHARE_GRADE)
tool.delete_dataset(
f[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_GRADE], today_str)
tool.merge_df_dataset(
f[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_GRADE], today_str,
grade_df)
f_share.close()
f.close()
console.write_tail()
return
def all_exec(omit_list):
"""
筛选出存在背离的股票
"""
# 筛选记录内容如下:
# 1. 月线macd趋势
# 2. 周线macd趋势
# 3. 日线macd趋势，是否背离，数值差距
# 4. 30min的macd趋势，是否背离，数值差距，震荡中枢数量
# 5. 5min的macd趋势，是否背离，数值差距，震荡中枢数量
console.write_head(conf.HDF5_OPERATE_SCREEN, conf.HDF5_RESOURCE_TUSHARE,
conf.STRATEGY_TREND_AND_REVERSE)
f = h5py.File(conf.HDF5_FILE_SHARE, 'a')
filter_df = tool.init_empty_df(_ini_filter_columns())
for code_prefix in f:
if code_prefix in omit_list:
continue
for code in f[code_prefix]:
code_group_path = '/' + code_prefix + '/' + code
if f.get(code_group_path) is None:
console.write_blank()
console.write_msg(code + "的tushare数据不存在")
continue
# 忽略停牌、退市、无法获取的情况
if f[code_prefix][code].attrs.get(
conf.HDF5_BASIC_QUIT
) is not None or f[code_prefix][code].attrs.get(
conf.HDF5_BASIC_ST) is not None:
console.write_blank()
console.write_msg(code + "已退市或停牌")
continue
try:
code_dict = code_exec(f, code)
if code_dict is None:
console.write_pass()
continue
else:
console.write_exec()
filter_df = filter_df.append(code_dict, ignore_index=True)
except Exception as er:
console.write_msg("[" + code + "]" + str(er))
f.close()
f_screen = h5py.File(conf.HDF5_FILE_SCREEN, 'a')
if f_screen.get(conf.STRATEGY_TREND_AND_REVERSE) is None:
f_screen.create_group(conf.STRATEGY_TREND_AND_REVERSE)
if f_screen[conf.STRATEGY_TREND_AND_REVERSE].get(
conf.SCREEN_SHARE_FILTER) is None:
f_screen[conf.STRATEGY_TREND_AND_REVERSE].create_group(
conf.SCREEN_SHARE_FILTER)
today_str = tradetime.get_today()
tool.delete_dataset(
f_screen[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_FILTER],
today_str)
tool.merge_df_dataset(
f_screen[conf.STRATEGY_TREND_AND_REVERSE][conf.SCREEN_SHARE_FILTER],
today_str, filter_df)
f_screen.close()
console.write_blank()
console.write_tail()
return
