def calc_and_save_stock_bias_quantile():
observation_period = 250
win_lens = [20, 60, 120]
session = DBSession()
sec_codes = get_sse50_stk_codes()
for sec_code in sec_codes:
bias_df = get_bias_eod(sec_code, is_index=False)
bias_df = bias_df.iloc[-observation_period:].copy() # 观察期内所有bias。 注意，sort会改变原来的值
last_bias_quantile_dict = dict()
last_trade_date = bias_df.index[-1]
for win_len in win_lens:
bias_df['value_%d' % win_len] = bias_df['value_%d' % win_len].abs() # bias全部转成 非负数
last_bias = bias_df['value_%d' % win_len][-1]
bias_arr = bias_df['value_%d' % win_len].values
bias_arr.sort() # 注意，会改变bias_arr本来的内容
if np.isnan(last_bias):
raise ValueError('最新数据是nan，检查数据')
last_bias_idx = np.argwhere(bias_arr == last_bias)[0][0]
last_bias_quantile = last_bias_idx / observation_period
last_bias_quantile_dict[win_len] = last_bias_quantile # 分位数
new_obj = AShareBiasQuantile()
new_obj.sec_code = sec_code
new_obj.trade_date = last_trade_date
new_obj.value_20 = trans_number_to_float(last_bias_quantile_dict[20])
new_obj.value_60 = trans_number_to_float(last_bias_quantile_dict[60])
new_obj.value_120 = trans_number_to_float(last_bias_quantile_dict[120])
new_obj.observation_period = observation_period
session.add(new_obj)
session.commit()
session.close()
def calc_and_save_rolling_alpha(daily_alpha_df):
# 注意：这里修改窗口长度
alpha_rolling_unit_value_df_20 = (1 + daily_alpha_df).rolling(20).apply(
calc_prod, raw=True)
alpha_rolling_unit_value_df_60 = (1 + daily_alpha_df).rolling(60).apply(
calc_prod, raw=True)
# save_df_into_db(alpha_rolling_unit_value_df, AShareAlpha)
df_20 = alpha_rolling_unit_value_df_20
mat_20 = df_20.values
mat_60 = alpha_rolling_unit_value_df_60.values
trade_dates = df_20.index
sec_codes = df_20.columns
session = DBSession()
for i in range(len(trade_dates)):
for j in range(len(sec_codes)):
trade_date = trade_dates[i]
sec_code = sec_codes[j]
new_obj = AShareAlpha()
new_obj.sec_code = sec_code
new_obj.trade_date = trade_date
new_obj.value_20 = trans_number_to_float(mat_20[i, j])
new_obj.value_60 = trans_number_to_float(mat_60[i, j])
session.add(new_obj)
session.commit()
session.close()
def calc_and_save_alpha_quantile():
observation_period = 250
win_len = 20
session = DBSession()
sec_codes = get_sse50_stk_codes()
for sec_code in sec_codes:
value_df = get_a_share_alpha(sec_code)
value_df = value_df.iloc[-observation_period:].copy() # 注意，sort会改变原来的值
last_trade_date = value_df.index[-1]
last_value = value_df['value_%d' % win_len][-1]
value_arr = value_df['value_%d' % win_len].values
value_arr = np.array([x for x in value_arr if x > 1e-6])
value_arr.sort() # 注意，会改变bias_arr本来的内容
if np.isnan(last_value):
raise ValueError('最新数据是nan，检查数据')
last_value_idx = np.argwhere(value_arr == last_value)[0][0]
last_value_quantile = last_value_idx / observation_period
new_obj = AShareAlphaQuantile()
new_obj.sec_code = sec_code
new_obj.trade_date = last_trade_date
new_obj.value_20 = trans_number_to_float(last_value_quantile)
new_obj.observation_period = observation_period
session.add(new_obj)
session.commit()
session.close()
def calc_and_save_stock_pb_quantile():
observation_period = 750
session = DBSession()
sec_codes = get_sse50_stk_codes()
for sec_code in sec_codes:
value_df = get_a_share_fin_pit(sec_code)
value_df = value_df.iloc[-observation_period:].copy() # sort会改变原来的值
last_trade_date = value_df.index[-1]
last_value = value_df['pb'][-1]
value_arr = value_df['pb'].values
value_arr.sort() # 注意，sort会改变bias_arr本来的内容
if np.isnan(last_value):
raise ValueError('最新数据是nan，检查数据')
last_value_idx = np.argwhere(value_arr == last_value)[0][0]
last_value_quantile = last_value_idx / observation_period
new_obj = ASharePbQuantile()
new_obj.sec_code = sec_code
new_obj.trade_date = last_trade_date
new_obj.val = trans_number_to_float(last_value_quantile)
new_obj.observation_period = observation_period
session.add(new_obj)
session.commit()
session.close()
def init_a_share_eod_prices():
session = DBSession()
years = [2016, 2017, 2018, 2019]
close_dfs = []
pct_chg_dfs = []
for year in years:
close_df = pd.read_csv('csv/{year}_close.csv'.format(year=year),
index_col='Unnamed: 0')
close_dfs.append(close_df)
pct_chg_df = pd.read_csv('csv/{year}_pct_chg.csv'.format(year=year),
index_col='Unnamed: 0')
pct_chg_dfs.append(pct_chg_df)
close_df = pd.concat(close_dfs)
pct_chg_df = pd.concat(pct_chg_dfs)
adjfactor_df = pd.read_csv('csv/cpnt_adjfactor.csv', index_col='DateTime')
sec_codes = close_df.columns
close_df.index = pd.to_datetime(close_df.index)
trade_dates = close_df.index.date
close_mat = close_df.values
pct_chg_mat = pct_chg_df.values
adjfactor_mat = adjfactor_df.values
for i in range(len(trade_dates)):
for j in range(len(sec_codes)):
trade_date = trade_dates[i]
sec_code = sec_codes[j]
new_obj = AShareEodPrice()
new_obj.sec_code = sec_code
new_obj.trade_date = trade_date
new_obj.close = trans_number_to_float(close_mat[i, j])
new_obj.pct_chg = trans_number_to_float(pct_chg_mat[i, j])
new_obj.adjfactor = trans_number_to_float(adjfactor_mat[i, j])
session.add(new_obj)
session.commit()
session.close()
def init_a_share_fin_pit():
session = DBSession()
years = [2016, 2017, 2018, 2019]
for year in years:
fa_bps_df = pd.read_csv('csv/{year}_fa_bps.csv'.format(year=year),
index_col='Unnamed: 0')
pe_ttm_df = pd.read_csv('csv/{year}_pe_ttm.csv'.format(year=year),
index_col='Unnamed: 0')
close_df = pd.read_csv('csv/{year}_close.csv'.format(year=year),
index_col='Unnamed: 0')
pb_df = close_df / fa_bps_df
sec_codes = close_df.columns
close_df.index = pd.to_datetime(close_df.index)
trade_dates = close_df.index.date
pb_mat = pb_df.values
pe_ttm_mat = pe_ttm_df.values
fa_bps_mat = fa_bps_df.values
for i in range(len(trade_dates)):
for j in range(len(sec_codes)):
trade_date = trade_dates[i]
sec_code = sec_codes[j]
new_obj = AShareFinPit()
new_obj.sec_code = sec_code
new_obj.trade_date = trade_date
new_obj.fa_bps = trans_number_to_float(fa_bps_mat[i, j])
new_obj.pe_ttm = trans_number_to_float(pe_ttm_mat[i, j])
new_obj.pb = trans_number_to_float(pb_mat[i, j])
session.add(new_obj)
session.commit()
session.close()
def save_df_into_db(df, clazz):
mat = df.values
trade_dates = df.index
sec_codes = df.columns
session = DBSession()
for i in range(len(trade_dates)):
for j in range(len(sec_codes)):
trade_date = trade_dates[i]
sec_code = sec_codes[j]
new_obj = clazz()
new_obj.sec_code = sec_code
new_obj.trade_date = trade_date
new_obj.value_20 = trans_number_to_float(mat[i, j])
session.add(new_obj)
session.commit()
session.close()
def save_indicators(indicators):
session = DBSession()
new_obj = AShareTalib()
new_obj.sec_code = indicators['sec_code']
new_obj.trade_date = indicators['trade_date']
new_obj.freq = 'D'
# ma, ema
new_obj.ma5 = trans_number_to_float(indicators['ma'][0])
new_obj.ma10 = trans_number_to_float(indicators['ma'][1])
new_obj.ma20 = trans_number_to_float(indicators['ma'][2])
new_obj.ma60 = trans_number_to_float(indicators['ma'][3])
new_obj.ma120 = trans_number_to_float(indicators['ma'][4])
new_obj.ema5 = trans_number_to_float(indicators['ema'][0])
new_obj.ema10 = trans_number_to_float(indicators['ema'][1])
new_obj.ema20 = trans_number_to_float(indicators['ema'][2])
new_obj.ema60 = trans_number_to_float(indicators['ema'][3])
new_obj.ema120 = trans_number_to_float(indicators['ema'][4])
# macd, rsi, stoch, adx, cci, stochrsi, uo, roc, sar
new_obj.macd = trans_number_to_float(indicators['macd'])
new_obj.rsi = trans_number_to_float(indicators['rsi'])
new_obj.cci = trans_number_to_float(indicators['cci'])
new_obj.adx = trans_number_to_float(indicators['adx'])
new_obj.stochrsi = trans_number_to_float(indicators['stochrsi'])
new_obj.stoch = trans_number_to_float(indicators['stoch'])
new_obj.uo = trans_number_to_float(indicators['uo'])
new_obj.roc = trans_number_to_float(indicators['roc'])
new_obj.sar = trans_number_to_float(indicators['sar'])
session.add(new_obj)
session.commit()
session.close()