def get_adjust_point(startdate='2010-01-08', enddate=None, label_period=1):
# 加载时间轴数据
if not enddate:
yesterday = (datetime.now() - timedelta(days=1))
enddate = yesterday.strftime("%Y-%m-%d")
index = base_trade_dates.load_trade_dates(startdate, end_date=enddate)
index = index[index.td_type >= 8].index
#index = DBData.trade_date_index(startdate, end_date=enddate)
if label_period > 1:
label_index = index[::label_period]
if index.max() not in label_index:
label_index = label_index.insert(len(label_index), index.max())
else:
label_index = index
# label_index = pd.DatetimeIndex([
# '2010-01-08',
# '2010-07-09',
# '2011-01-07',
# '2011-07-08',
# '2012-01-13',
# '2012-07-20',
# '2013-01-25',
# '2013-08-02',
# '2014-01-30',
# '2014-08-01',
# '2015-01-30',
# '2015-07-31',
# '2016-01-29',
# '2016-08-05',
# '2016-11-05',
# ])
return label_index
def trade_date(begin_date = None, end_date = None, lookback = None):
if ATradeDate.__a_trade_date is None:
ATradeDate.__a_trade_date = base_trade_dates.load_trade_dates().sort_index()
td_date = ATradeDate.__a_trade_date.copy()
if begin_date is not None:
if lookback is not None:
tmp_td_date = td_date[td_date.index <= begin_date]
td_date = td_date[td_date.index >= tmp_td_date.index[-1 * lookback]]
else:
td_date = td_date[td_date.index >= begin_date]
if end_date is not None:
td_date = td_date[td_date.index <= end_date]
return td_date.index
def month_trade_date(begin_date = None, end_date = None, lookback = None):
if ATradeDate.__a_trade_date is None:
ATradeDate.__a_trade_date = base_trade_dates.load_trade_dates().sort_index()
td_date = ATradeDate.__a_trade_date.copy()
yesterday = td_date.iloc[[-2], :]
td_date = td_date[(td_date.td_type & 0x08) > 0]
if not yesterday.index in td_date.index:
td_date = pd.concat([td_date, yesterday])
if begin_date is not None:
if lookback is not None:
tmp_td_date = td_date[td_date.index <= begin_date]
td_date = td_date[td_date.index >= tmp_td_date.index[-1 * lookback]]
else:
td_date = td_date[td_date.index >= begin_date]
if end_date is not None:
td_date = td_date[td_date.index <= end_date]
return td_date.index
def fc_update_bflayer_nav(ctx, optid):
engine = database.connection('asset')
Session = sessionmaker(bind=engine)
session = Session()
sql1 = session.query(factor_cluster_asset.fc_asset_id).filter(
factor_cluster_asset.globalid == optid)
asset_ids = [asset[0] for asset in sql1.all()]
assets = {}
for asset_id in asset_ids:
print asset_id
assets[asset_id] = load_nav_series(asset_id)
start_date = '2010-01-01'
# end_date = '2018-12-01'
end_date = datetime.datetime.now().strftime('%Y-%m-%d')
# trade_dates = DBData.trade_dates(start_date = '2010-01-01', end_date = '2018-12-01')
# trade_dates = base_trade_dates.load_trade_dates(begin_date = '2010-01-01', end_date = '2018-12-01').index
trade_dates = base_trade_dates.load_trade_dates(begin_date='2010-01-01',
end_date=end_date).index
df_assets = pd.DataFrame(assets)
df_assets = df_assets[df_assets.index >= start_date]
df_assets = df_assets[df_assets.index <= end_date]
df_assets = df_assets.reindex(trade_dates).dropna()
df_ret = df_assets.pct_change().dropna()
sql2 = session.query(distinct(barra_stock_factor_valid_factor.trade_date))
select_dates = [select_date[0] for select_date in sql2.all()]
select_dates = select_dates[20:]
sql3 = session.query(
barra_stock_factor_valid_factor.trade_date,
barra_stock_factor_valid_factor.bf_layer_id,
).statement
valid_factors = pd.read_sql(sql3,
session.bind,
index_col=['trade_date'],
parse_dates=True)
df_new = pd.DataFrame(
columns=['date', 'nav', 'fc_cluster_id', 'factor_selected_date'])
for date in select_dates:
print date
bf_ids = valid_factors.loc[date].values.ravel()
factors_ret = df_ret.loc[:, bf_ids]
layer_ret = factors_ret.mean(1)
layer_nav = (1 + layer_ret).cumprod()
layer_nav = layer_nav.to_frame(name='nav')
layer_nav.index.name = 'date'
layer_nav['fc_cluster_id'] = '%s.%d' % (optid, 2)
layer_nav = layer_nav.reset_index()
layer_nav['factor_selected_date'] = date
df_new = pd.concat([df_new, layer_nav])
df_new['globalid'] = optid
df_new = df_new.set_index(
['globalid', 'fc_cluster_id', 'date', 'factor_selected_date'])
db = database.connection('asset')
metadata = MetaData(bind=db)
t = Table('factor_cluster_nav', metadata, autoload=True)
df_old = asset_factor_cluster_nav.load_nav(optid)
database.batch(db, t, df_new, df_old)
def fc_update_nav(ctx, optid):
engine = database.connection('asset')
Session = sessionmaker(bind=engine)
session = Session()
sql1 = session.query(factor_cluster_asset.fc_asset_id).filter(
factor_cluster_asset.globalid == optid)
asset_ids = [asset[0] for asset in sql1.all()]
assets = {}
for asset_id in asset_ids:
assets[asset_id] = load_nav_series(asset_id)
start_date = '2010-01-01'
end_date = '2018-12-01'
# trade_dates = DBData.trade_dates(start_date = '2010-01-01', end_date = '2018-12-01')
trade_dates = base_trade_dates.load_trade_dates(
begin_date='2010-01-01', end_date='2018-12-01').index
df_assets = pd.DataFrame(assets)
df_assets = df_assets[df_assets.index >= start_date]
df_assets = df_assets[df_assets.index <= end_date]
df_assets = df_assets.reindex(trade_dates).dropna()
df_ret = df_assets.pct_change().dropna()
df_layer_result = pd.read_csv('df_layer_result.csv',
index_col=0,
parse_dates=True)
dates = df_layer_result.index.drop_duplicates()
df_new = pd.DataFrame(
columns=['date', 'nav', 'fc_cluster_id', 'factor_selected_date'])
for date in dates:
print date
df = df_layer_result.loc[date]
layers = np.unique(df.layer)
df_layer = pd.DataFrame(columns=['date', 'nav', 'fc_cluster_id'])
for layer in layers:
factors = df[df.layer == layer]
factors = factors.factor.values
factors_ret = df_ret.loc[:, factors]
layer_ret = factors_ret.mean(1)
layer_nav = (1 + layer_ret).cumprod()
layer_nav = layer_nav.to_frame(name='nav')
layer_nav.index.name = 'date'
layer_nav['fc_cluster_id'] = optid + '.%d' % layer
layer_nav = layer_nav.reset_index()
df_layer = pd.concat([df_layer, layer_nav])
df_layer['factor_selected_date'] = date
df_new = pd.concat([df_new, df_layer])
df_new['globalid'] = optid
df_new = df_new.set_index(
['globalid', 'fc_cluster_id', 'date', 'factor_selected_date'])
db = database.connection('asset')
metadata = MetaData(bind=db)
t = Table('factor_cluster_nav', metadata, autoload=True)
df_old = asset_factor_cluster_nav.load_nav(optid)
database.batch(db, t, df_new, df_old)