def plot_send_data(raw_df, subject, text=''):
raw_df.plot(legend=True)
plt.savefig(f'{figure_save_path}/{subject}.png')
plt.close()
to = ['*****@*****.**']
filepath = [f'{figure_save_path}/{subject}.png']
send_email.send_email(text, to, filepath, subject)
def savfig_send(subject='tmp', text='', to=None, filepath=None):
target_save_path = f'{figure_save_path}/{subject}.png'
if to is None:
to = ['*****@*****.**']
if filepath is None:
filepath = [target_save_path]
plt.savefig(target_save_path)
send_email.send_email(text, to, filepath, subject)
plt.close()
def plot_send_result(pnl_df, sharpe_ratio, subject, text=''):
plt.figure(figsize=[16, 8])
plt.plot(pnl_df.index, pnl_df.cumsum(), label='sharpe_ratio={}'.format(sharpe_ratio))
plt.grid()
plt.legend()
plt.savefig(os.path.join(figure_save_path, '{}.png'.format(subject)))
plt.close()
to = ['*****@*****.**']
filepath = [os.path.join(figure_save_path, '{}.png'.format(subject))]
send_email.send_email(text, to, filepath, subject)
def plot_send_result_mul(pnl_df, subject, text=''):
assert type(pnl_df) == pd.DataFrame
pnl_num = len(pnl_df.columns)
plt.figure(figsize=[16, 8 * pnl_num])
for i, col in enumerate(pnl_df.columns):
ax = plt.subplot(pnl_num, 1, i + 1)
ax.plot(pnl_df[col].index, pnl_df[col].cumsum(), label=f'{col}, sharpe_ratio={bt.AZ_Sharpe_y(pnl_df[col])}')
plt.grid()
plt.legend()
plt.savefig(os.path.join(figure_save_path, '{}.png'.format(subject)))
plt.close()
to = ['*****@*****.**']
filepath = [os.path.join(figure_save_path, '{}.png'.format(subject))]
send_email.send_email(text, to, filepath, subject)
def corr_test_fun(pnl_df, alpha_name):
sum_pnl_df = get_corr_matrix(cut_date=None)
sum_pnl_df_c = pd.concat([sum_pnl_df, pnl_df], axis=1)
corr_self = sum_pnl_df_c.corr()[[alpha_name]]
other_corr = get_all_pnl_corr(pnl_df, alpha_name)
print(other_corr)
self_corr = corr_self[corr_self > 0.7].dropna(axis=0)
print(self_corr)
if len(self_corr) >= 2 or len(other_corr) >= 2:
print('FAIL!')
send_email.send_email('FAIL!\n' + corr_self.to_html(),
['*****@*****.**'], [],
'[RESULT DEAL]' + alpha_name)
else:
print('SUCCESS!')
send_email.send_email('SUCCESS!\n' + corr_self.to_html(),
['*****@*****.**'], [],
'[RESULT DEAL]' + alpha_name)
print('______________________________________')
def config_test(main_model, config_name, result_file_name, cut_date):
config_set = pd.read_pickle(
f'/mnt/mfs/dat_whs/alpha_data/{config_name}.pkl')
config_data = config_set['factor_info']
sum_factor_df = pd.DataFrame()
for i in config_data.index:
fun_name, name1, name2, name3, buy_sell = config_data.loc[i]
print('***************************************************')
print('now {}\'s is running, key={}, {}, {}, {}'.format(
i, fun_name, name1, name2, name3))
mix_factor, con_in_c, con_out_c, ic_c, sp_u_c, sp_m_c, sp_d_c, pot_in_c, fit_ratio_c, leve_ratio_c, sp_in_c, \
sp_out_c, pnl_df_c = main_model.single_test(fun_name, name1, name2, name3)
# plot_send_result(pnl_df_c, bt.AZ_Sharpe_y(pnl_df_c), '{}, {}, {}, {}, {}'
# .format(fun_name, name1, name2, name3, buy_sell))
# print(con_in_c, con_out_c, ic_c, sp_u_c, sp_m_c, sp_d_c, pot_in_c, fit_ratio_c, leve_ratio_c, sp_out_c)
if buy_sell > 0:
sum_factor_df = sum_factor_df.add(mix_factor, fill_value=0)
else:
sum_factor_df = sum_factor_df.add(-mix_factor, fill_value=0)
sum_pos_df = main_model.deal_mix_factor(sum_factor_df).shift(2)
in_condition, out_condition, ic, sharpe_q_in_df_u, sharpe_q_in_df_m, sharpe_q_in_df_d, pot_in, \
fit_ratio, leve_ratio, sp_in, sharpe_q_out, pnl_df = filter_all(cut_date, sum_pos_df, main_model.return_choose,
if_return_pnl=True, if_only_long=False)
print(in_condition, out_condition, ic, sharpe_q_in_df_u, sharpe_q_in_df_m,
sharpe_q_in_df_d, pot_in, fit_ratio, leve_ratio, sp_in, sharpe_q_out)
sp = bt.AZ_Sharpe_y(pnl_df)
pnl_df.to_csv(f'/mnt/mfs/dat_whs/tmp_pnl_file/{result_file_name}.csv')
send_list = [
in_condition, out_condition, ic, sharpe_q_in_df_u, sharpe_q_in_df_m,
sharpe_q_in_df_d, pot_in, fit_ratio, leve_ratio, sp_in, sharpe_q_out
]
send_email.send_email(','.join([str(x) for x in send_list]),
['*****@*****.**'], [],
'[RESULT DEAL]' + result_file_name)
plot_send_result(pnl_df, bt.AZ_Sharpe_y(pnl_df),
'[RESULT DEAL]' + result_file_name)
return sum_pos_df, pnl_df, sp
def main(result_file_name, time_para_dict):
print(
'*******************************************************************************************************'
)
root_path = '/mnt/mfs/DAT_EQT'
# result_file_name = 'market_top_800plus_True_20181104_0237_hold_5__7'
# root_path = '/media/hdd1/DAT_EQT'
config_name = result_file_name
if_save = False
if_new_program = True
hold_time = int(result_file_name.split('hold')[-1].split('_')[1])
# 加载对应脚本
if result_file_name.split('_')[-1] == 'long':
script_num = result_file_name.split('_')[-2]
if_only_long = True
else:
script_num = result_file_name.split('_')[-1]
if_only_long = False
# script_num = '15'
print('script_num : ', script_num)
print('hold_time : ', hold_time)
loc = locals()
exec(
f'from work_whs.AZ_2018_Q2.factor_script.main_file import main_file_sector_{script_num} as mf'
)
mf = loc['mf']
time_para_dict = mf.time_para_dict
# from work_whs.AZ_2018_Q2.factor_script.main_file import main_file_sector_6 as mf
lag = 2
return_file = ''
sector_name, if_hedge = find_sector_name(result_file_name)
if_hedge = True
# sector_name = 'market_top_800plus_industry_10_15'
print(result_file_name)
print(sector_name)
result_path = '/mnt/mfs/dat_whs/result/result/{}.txt'.format(
result_file_name)
# #############################################################################
# 判断文件大小
if os.path.getsize(result_path):
data = pd.read_csv(result_path,
sep='|',
header=None,
error_bad_lines=False)
data.columns = [
'time_para', 'key', 'fun_name', 'name1', 'name2', 'name3',
'filter_fun_name', 'sector_name', 'con_in', 'con_out_1',
'con_out_2', 'con_out_3', 'con_out_4', 'ic', 'sp_u', 'sp_m',
'sp_d', 'pot_in', 'fit_ratio', 'leve_ratio', 'sp_in', 'sp_out_1',
'sp_out_2', 'sp_out_3', 'sp_out_4'
]
else:
return 0
filter_cond = data[['name1', 'name2', 'name3']] \
.apply(lambda x: not (('R_COMPANYCODE_First_row_extre_0.3' in set(x)) or
('return_p20d_0.2' in set(x)) or
('price_p120d_hl' in set(x)) or
('return_p60d_0.2' in set(x)) or
('wgt_return_p120d_0.2' in set(x)) or
('wgt_return_p20d_0.2' in set(x)) or
('log_price_0.2' in set(x)) or
('TVOL_row_extre_0.2' in set(x)) or
('TVOL_row_extre_0.2' in set(x)) or
('tab2_11_row_extre_0.3' in set(x)) or
('tab1_8_row_extre_0.3' in set(x)) or
('intra_dn_vol_col_score_row_extre_0.3' in set(x)) or
('intra_dn_vol_row_extre_0.3' in set(x)) or
('turn_p30d_0.24' in set(x)) or
('evol_p30d' in set(x))
# ('CMO_40_0' in set(x))
# ('ATR_40_0.2' in set(x))
# ('ADX_200_40_20' in set(x))
# ('ATR_140_0.2' in set(x))
), axis=1)
data = data[filter_cond]
para_adj_set_list = [{
'pot_in_num': 50,
'leve_ratio_num': 2,
'sp_in': 1.5,
'ic_num': 0.0,
'fit_ratio': 2
}, {
'pot_in_num': 40,
'leve_ratio_num': 2,
'sp_in': 1.5,
'ic_num': 0.0,
'fit_ratio': 2
}, {
'pot_in_num': 50,
'leve_ratio_num': 2,
'sp_in': 1,
'ic_num': 0.0,
'fit_ratio': 1
}, {
'pot_in_num': 50,
'leve_ratio_num': 1,
'sp_in': 1,
'ic_num': 0.0,
'fit_ratio': 2
}, {
'pot_in_num': 50,
'leve_ratio_num': 1,
'sp_in': 1,
'ic_num': 0.0,
'fit_ratio': 1
}, {
'pot_in_num': 40,
'leve_ratio_num': 1,
'sp_in': 1,
'ic_num': 0.0,
'fit_ratio': 1
}]
# para_adj_set_list_9 = [{'pot_in_num': 30, 'leve_ratio_num': 2, 'sp_in': 1.5, 'ic_num': 0.0, 'fit_ratio': 2},
# {'pot_in_num': 20, 'leve_ratio_num': 2, 'sp_in': 1.5, 'ic_num': 0.0, 'fit_ratio': 2},
# {'pot_in_num': 30, 'leve_ratio_num': 2, 'sp_in': 1, 'ic_num': 0.0, 'fit_ratio': 1},
# {'pot_in_num': 30, 'leve_ratio_num': 1, 'sp_in': 1, 'ic_num': 0.0, 'fit_ratio': 2},
# {'pot_in_num': 30, 'leve_ratio_num': 1, 'sp_in': 1, 'ic_num': 0.0, 'fit_ratio': 1},
# {'pot_in_num': 20, 'leve_ratio_num': 1, 'sp_in': 1, 'ic_num': 0.0, 'fit_ratio': 1}]
time_para = 'time_para_5'
print(time_para)
# #############################################################################
# 结果分析
print('结果分析')
survive_result = survive_ratio_test(data, para_adj_set_list)
if survive_result is None:
print(f'{result_file_name} not satisfaction!!!!!!!!')
return 0
else:
pass
print(hold_time)
#############################################################################
# 回测函数
if sector_name.startswith('market_top_300plus'):
if_weight = 1
ic_weight = 0
elif sector_name.startswith('market_top_300to800plus'):
if_weight = 0
ic_weight = 1
else:
if_weight = 0.5
ic_weight = 0.5
print('回测函数')
begin_date, cut_date, end_date, end_date, end_date, end_date = time_para_dict[
time_para]
main_model = mf.FactorTestSector(root_path, if_save, if_new_program,
begin_date, cut_date, end_date,
time_para_dict, sector_name, hold_time,
lag, return_file, if_hedge, if_only_long,
if_weight, ic_weight)
sum_pos_df, pnl_df = pos_sum_c(main_model, data, time_para,
result_file_name, **survive_result)
#############################################################################
# 生成config文件
config_create(main_model,
sector_name,
result_file_name,
config_name,
data,
time_para,
**survive_result,
n=5,
use_factor_num=40)
###########################################################################
# 测试config结果
begin_date, cut_date, end_date, end_date, end_date, end_date = time_para_dict[
time_para]
sum_pos_df, pnl_df, sp = config_test(main_model, config_name,
result_file_name, cut_date)
if sp < 2:
return 0
pnl_df.name = result_file_name
# #############################################################################
# 计算相关性
# pnl_df_CRTSECJUN
#
sum_pnl_df = get_corr_matrix(cut_date=None)
sum_pnl_df_c = pd.concat([sum_pnl_df, pnl_df], axis=1)
corr_self = sum_pnl_df_c.corr()[[result_file_name]]
print(corr_self)
print('______________________________________')
print(corr_self[corr_self > 0.7].dropna(axis=0))
if len(corr_self[corr_self > 0.7].dropna(axis=0)) >= 2:
print('FAIL!')
send_email.send_email('FAIL!\n' + pd.DataFrame(corr_self).to_html(),
['*****@*****.**'], [], result_file_name)
else:
print('SUCCESS!')
send_email.send_email('SUCCESS!\n' + pd.DataFrame(corr_self).to_html(),
['*****@*****.**'], [], result_file_name)
print('______________________________________')
return 0