def get_samples(mode=0, *arge, **kwargs):
''' mode 对应的测试集
0. 每种规模分别取前N，默认000300-沪深300，000001-上证综指，000905-中证500，399005-深证中小板指,399006-创业板指
1. 每个行业龙头，主要市值排行
2. 流动性看：成交量活跃度，按换手率，和成交量日间日内波动
3. 异动股，日间（短期）、日内
4. 风格： 动量、反转
5. 对每种指数，300，500
'''
total_num = kwargs.get('total_num') or 100
start_date = "20191201"
end_date = "20191231"
if mode == 0:
# tickers = ['000300','000001','000905','399005','399006','399001']
tickers = ['399005', '399006', '399001']
ticker_to_exchangecd = {
'000300': 'XSHG',
'000001': 'XSHG',
'000905': 'XSHG',
'399005': 'XSHE',
'399006': 'XSHE',
'399001': 'XSHE'
}
ret = {}
_num = int(total_num / len(tickers))
for _ticker in tickers:
df = DataAPI.mIdxCloseWeightGet(secID=u"",
ticker=_ticker,
beginDate=start_date,
endDate=end_date,
field=["consID", "weight"],
pandas="1").sort_values(
by='weight', ascending=False)
# TODO choose the top weights coons, then the correlations with the becnchmark is high, change to sample num
# randomly
sec_ids = df['consID'][:_num]
ret.update({
'{0}.{1}'.format(_ticker, ticker_to_exchangecd.get(_ticker)):
sec_ids
})
return ret
def get_samples(mode=0, *arge, **kwargs):
''' mode 对应的测试集
0. 每种规模分别取前N，默认000300-沪深300，000001-上证综指，000905-中证500，399005-深证中小板指,399006-创业板指
1. 每个行业龙头，主要市值排行
2. 流动性看：成交量活跃度，按换手率，和成交量日间日内波动, 现为换手率，对设置的市场
3. 异动股，日间（短期）、日内
4. 风格： 动量、反转
5. 对每种指数，300，500
'''
total_num = kwargs.get('total_num') or 100
start_date = kwargs.get('start_date') or "20191201"
end_date = kwargs.get('end_date') or "20191231"
# ['000300', '000001', '000905', '399005', '399006', '399001'] #sh and sz market
# ['399005', '399006', '399001'] #sz market
tickers = kwargs.get('mkt_tickers') or ['399005', '399006', '399001']
ticker_to_exchangecd = {
'000300': 'XSHG',
'000001': 'XSHG',
'000905': 'XSHG',
'399005': 'XSHE',
'399006': 'XSHE',
'399001': 'XSHE'
}
if mode == 0:
ret = {}
_num = int(total_num / len(tickers))
for _ticker in tickers:
df = DataAPI.mIdxCloseWeightGet(secID=u"",
ticker=_ticker,
beginDate=start_date,
endDate=end_date,
field=["consID", "weight"],
pandas="1").sort_values(
by='weight', ascending=False)
# TODO choose the top weights coons, then the correlations with the becnchmark is high, change to sample num
# randomly
sec_ids = df['consID'][:_num]
ret.update({
'{0}.{1}'.format(_ticker, ticker_to_exchangecd.get(_ticker)):
sec_ids
})
elif mode == 2:
ret = {}
_num = int(total_num / len(tickers))
for _ticker in tickers:
df = DataAPI.mIdxCloseWeightGet(secID=u"",
ticker=_ticker,
beginDate=start_date,
endDate=end_date,
field=["consID", "weight"],
pandas="1").sort_values(
by='weight', ascending=False)
# TODO choose the top weights coons, then the correlations with the becnchmark is high, change to sample num
# randomly
cons_ids = df['consID']
mkt_df = DataAPI.MktEqudGet(secID=cons_ids,
beginDate=start_date,
endDate=end_date,
isOpen=1,
pandas="1")
df_agg = mkt_df.groupby('secID').agg({'turnoverRate': ['mean']})
flatten_columns = [
'{0}_{1}'.format(item[0], item[1]) for item in df_agg.columns
]
df_agg.columns = flatten_columns
df_agg = df_agg.reset_index()
df_agg.sort_values(by='turnoverRate_mean', ascending=False)
sec_ids = df_agg['secID'][:_num]
ret.update({
'{0}.{1}'.format(_ticker, ticker_to_exchangecd.get(_ticker)):
sec_ids
})
return ret
