def period_return_maker_(chooseday_r, factor_value, indicator):
date_r2 = month(chooseday_r, change_month=-12)
date = chooseday_r
if len(date) <= 8:
date_r = date[0:4] + '-' + date[4:6] + '-' + date[6:8]
else:
date_r = date
# 创建某时刻截面指标df
df_values = pd.DataFrame(columns=['date', 'code', 'factor_value'],
index=range(len(factor_value)))
df_values['date'] = date
df_values['code'] = factor_value['code']
df_values['factor_value'] = factor_value[indicator] # 指标选择
# 对指标进行标准化
df_values['factor_value'] = (df_values['factor_value'] -
df_values['factor_value'].min()) \
/ (df_values['factor_value'].max() -
df_values['factor_value'].min()) # 标准化
# 创建该段时间所有股票的收益率pool
returns = get_data_panel("huangwei",
"huangwei",
list(df_values['code']),
date_r2,
date_r,
collection='stock_daily')
df_price = pd.DataFrame(columns=['date', 'code'],
index=range(len(df_values)))
df_price['date'] = date
df_price['code'] = df_values['code']
df_price['cum_return'] = 0
# 计算每只股票在这段时间的累计收益，并填入cum_return中
def caculate_cum_return(code):
one_stock = returns[returns['code'] == code].reset_index(drop=True)
unique_index = list(one_stock['date'].drop_duplicates(
keep='first', inplace=False).index)
one_stock = one_stock.iloc[unique_index].sort_values(by='date')
one_stock['rtn'] += 1
one_stock['c_rtn'] = one_stock['rtn'].cumprod()
return one_stock['c_rtn'].iloc[-1] - one_stock['c_rtn'].iloc[0]
df_price['cum_return'] = df_price['code'].apply(caculate_cum_return)
# 合并收盘价以及指标值数据，计算两者截面相关性
df_all = df_price.merge(df_values[['code', 'factor_value']],
how='inner',
on='code')
# 剔除不匹配的股票
df_all = df_all.dropna(axis=0, how='any').reset_index(drop=True)
ic, p = spearman_cor(df_all['cum_return'].values,
df_all['factor_value'].values)
print('IC值计算完毕')
return ic, p
def _open_convert_database(self):
all_data = get_data_panel("huangwei",
"huangwei",
self.symbol_list,
'2017-03-01',
'2019-08-01',
collection='stock_daily')
comb_index = None
for s in self.symbol_list:
one_stock = all_data[all_data['code'] == s][[
'date', 'open', 'high', 'low', 'close', 'vol', 'rtn'
]]
one_stock = one_stock.rename(columns={
"date": "datetime",
"vol": "volume"
})
one_stock['adj_close'] = 0
one_stock_arr = one_stock.values
for i in range(len(one_stock_arr)):
if i == 0:
one_stock_arr[i, 7] = one_stock_arr[i, 4] \
* (1 + one_stock_arr[i, 6])
else:
one_stock_arr[i, 7] = one_stock_arr[i - 1, 7] \
* (1 + one_stock_arr[i, 6])
one_stock = pd.DataFrame(one_stock_arr,
columns=one_stock.columns,
index=range(len(one_stock)))
del one_stock['rtn']
# 将datatime作为index，并改为datatime格式
one_stock['datetime'] = pd.to_datetime(one_stock['datetime'])
one_stock.set_index(["datetime"], inplace=True)
one_stock = one_stock.fillna(method='ffill')
# ffill：用前一个非缺失值去填充该缺失值（向后填充）
# bfill：用下一个非缺失值填充该缺失值（向前填充）
self.symbol_data[s] = one_stock.sort_values(by="datetime",
ascending=True)
# Combine the index to pad forward values
if comb_index is None:
comb_index = self.symbol_data[s].index
else:
comb_index.union(self.symbol_data[s].index)
# 这里对于不同股票的日期会不断叠加，直到得到一列所有股票里面最长的
# Set the latest symbol_data to None
self.latest_symbol_data[s] = []
# Reindex the dataframes
for s in self.symbol_list:
self.symbol_data[s] = self.symbol_data[s].reindex(
index=comb_index, method='pad').iterrows()
def get_data_(cal_day, indicator):
# 读取所有股票信息表
stock_info = pd.read_excel(
'C:/Users/Lenovo/PycharmProjects/my_project/frjj_codes/stock_info_20190801.xlsx'
)
stock_info = stock_info.dropna(axis=0, how='any').reset_index(drop=True)
# 与指数成分股取交集
index_content_now = index_content[index_content['date'] == list(
index_content['date'][index_content['date'] >= cal_day])
[0]].reset_index(drop=True)
index_content_now.rename(columns={'stock_code': '证券代码'}, inplace=True)
stock_info = stock_info.merge(index_content_now['证券代码'],
on='证券代码',
how='inner')
# 读取申万一级行业划分
sw1 = sw_info_(cal_day, "sw1")
sw1.rename(columns={'sw_name': 'sw1_name'}, inplace=True)
# 读取申万三级行业划分
sw3 = sw_info_(cal_day, "sw3")
sw3.rename(columns={'sw_name': 'sw3_name'}, inplace=True)
all = pd.DataFrame(columns=['code', 'code_name'],
index=range(len(stock_info)))
all['code'] = stock_info['证券代码']
all['code_name'] = stock_info['证券简称']
all = all.merge(sw1[['code', 'sw1_code', 'sw1_name']],
on='code',
how='left')
all = all.merge(sw3[['code', 'sw3_code', 'sw3_name']],
on='code',
how='left')
# 使cal_day适应api
if len(cal_day) == 8:
cal_day = cal_day[0:4] + '-' + cal_day[4:6] + '-' + cal_day[6:8]
else:
pass
# 直接从数据库收集股票的流通市值（或总市值）
data = get_data_panel("huangwei",
"huangwei",
list(all['code']),
forwardchangeday2(cal_day),
forwardchangeday2(cal_day),
collection='stock_indicator')
# 用total_mv替换原表的cir_mv
all = all.merge(data[['code', indicator]], on='code', how='left')
all['cir_mv'] = all[indicator]
del all[indicator]
return all
def weight_maker_(chooseday_r, weight_origin, how=0):
user = "******"
passwd = "huangwei"
weight = weight_origin
if how == 0: # 等权
weight['weight_adj_sw1'] = 1 / len(weight_origin)
print('持仓赋权方式为：等权')
elif how == 1: # 按市值加权，取出这天股票组合的市值信息
if 'total_mv' not in weight.columns:
info_all = get_data_panel(user,
passwd,
list(weight['code']),
chooseday_r,
chooseday_r,
collection='stock_indicator')
weight = weight.merge(info_all[['code', 'total_mv']],
how='inner',
on='code')
else:
pass
weight_sum = sum(weight['total_mv'])
weight['weight_adj_sw1'] = weight['total_mv'] / weight_sum
del weight['total_mv']
print('持仓赋权方式为：市值加权')
elif how == 2: # 按过去一年收益率波动率倒数
# 获取当下日期过去一年的收益率数据
info_all = get_data_panel_backwards(user,
passwd,
list(weight['code']),
chooseday_r,
back_years=0,
back_months=12,
collection='stock_daily')
info_all = info_all.rename(columns={'code': 'codes'})
# 分组统计标准差
grouped = info_all.groupby(by=info_all['codes']).std()
grouped['code'] = grouped.index
weight = weight.merge(grouped[['code', 'rtn']], how='inner', on='code')
weight['rtn'] = 1 / weight['rtn']
weight['rtn'] = weight['rtn'].ffill() # 缺失值延续前值
weight['weight_sum'] = sum(weight['rtn'])
weight['weight_adj_sw1'] = weight['rtn'] / weight['weight_sum']
print('持仓赋权方式为：波动率倒数加权')
return weight
def get_data_(cal_day, indicator):
# 与指数成分股取交集
index_content_now = index_content[index_content['date'] == list(
index_content['date'][index_content['date'] >= cal_day])
[0]].reset_index(drop=True)
index_content_now.rename(columns={'stock_code': '证券代码'}, inplace=True)
stock_info_ = stock_info.merge(index_content_now['证券代码'],
on='证券代码',
how='inner')
# 取当天的申万行业划分
sw_ = sw[sw['date'] == cal_day].reset_index(drop=True)
all = pd.DataFrame(columns=['code', 'code_name'],
index=range(len(stock_info_)))
all['code'] = stock_info_['证券代码']
all['code_name'] = stock_info_['证券简称']
all = all.merge(sw_[[
'code', 'sw1_code', 'sw1_name', 'sw2_code', 'sw2_name', 'sw3_code',
'sw3_name'
]],
on='code',
how='left')
# 使cal_day适应api
if len(cal_day) == 8:
cal_day = cal_day[0:4] + '-' + cal_day[4:6] + '-' + cal_day[6:8]
else:
pass
# 直接从数据库收集股票的流通市值（或总市值）
data = get_data_panel("huangwei",
"huangwei",
list(all['code']),
forwardchangeday2(cal_day),
forwardchangeday2(cal_day),
collection='stock_indicator')
# 用total_mv替换原表的cir_mv
all = all.merge(data[['code', indicator]], on='code', how='left')
all['cir_mv'] = all[indicator]
del all[indicator]
return all
def position_init_(stock_pool, changeday, total_capitals):
# 持仓矩阵初始化
position = pd.DataFrame(columns=['date', 'code', 'weight', 'account',
'open_price',
'lots', 'account_surplus', 'close_price'],
index=range(len(stock_pool)))
position['code'] = stock_pool['code']
position['weight'] = stock_pool['weight_adj_sw1']
position['date'] = changeday
position['account'] = total_capitals[-1] * position['weight']
# 计算每个股票的应买手数
price = get_data_panel("huangwei", "huangwei", list(position['code']),
changeday, changeday,
collection='stock_daily').drop_duplicates(['code'])
if 'open' in position.columns:
del position['open']
if 'rtn' in position.columns:
del position['rtn']
position = position.merge(price[['code', 'open', 'rtn']],
on='code', how='inner')
position['open_price'] = position['open']
position['close_price'] = position['open_price'] * (1 + position['rtn'])
position['lots'] = position['account'] / (100 * position['open_price'])
# 对于获取不到开盘价的股票，可买手数一律赋0
position['lots'] = position['lots'].fillna(0)
# 为了方便计算剩余资金，创造facked字段，开盘价为nan的赋为0
position['open_price_faked'] = position['open_price'].fillna(0)
position['lots'] = position['lots'].apply(lambda x: math.floor(x))
position['account_realized'] = 100 * position['open_price_faked'] * position['lots']
position['account_surplus'] = position['account'] - position['account_realized']
return position
'cir_mv_y': 'industry_total_sw1'
},
inplace=True)
# 申万一级行业的比重*行业内部个股占整个行业的比重
result['weight_adj_sw1'] = result['industryweight_in_all'] * result[
'weight_in_industry_sw1']
result['weight_adj_sw1'] = result['weight_adj_sw1'] / result[
'weight_adj_sw1'].sum()
# 计算股票池的净值曲线
# 获得每只股票在回测期的日收益率（除权调整），累积日收益率数据
begin_date = '20160104'
end_date = "20190628"
data = get_data_panel("huangwei", "huangwei", list(result_top100mv['code']),
begin_date, end_date)
data_unlimited = get_data_panel("huangwei", "huangwei", list(result['code']),
begin_date, end_date)
dates = get_tradingdays("huangwei", "huangwei", begin_date, end_date)
dates = dates.rename(columns={'tradingdays': 'date'})
for i in range(len(dates)):
dates['date'].iloc[i] = dates['date'].iloc[i][0:4] + '-' + \
dates['date'].iloc[i][4:6] + '-' + dates['date'].iloc[i][6:8]
df_group_data_limited = dates
df_group_data_unlimited = dates
# 限定在沪深300成分股内
for code in list(result_top100mv['code']):
temp = data[data['code'] == code].reset_index(drop=True)
unique_index = list(temp['date'].drop_duplicates(keep='first',
if result_export:
# 输出持仓信息
export_result_(position, changeday, result_path, file_name1)
# 记录当天的损益
pnls, total_capitals = return_maker_(position, pnls, total_capitals,
transaction_cost_rate)
# last_stock_pool = stock_pool_class
elif changeday != change_days[0] and changeday in change_days: # 调仓日
# 更新开盘价数据
price = get_data_panel("huangwei",
"huangwei",
list(position['code']),
changeday,
changeday,
collection='stock_daily').drop_duplicates(
['code'])
# 保存过去的持仓情况
last_position = position
if 'open' in last_position.columns:
del last_position['open']
if 'rtn' in last_position.columns:
del last_position['rtn']
last_position = last_position.merge(price[['code', 'open', 'rtn']],
on='code',
how='inner')
last_position['open_price'] = last_position['open']
last_position['close_price'] = last_position['open_price'] \
* (1 + last_position['rtn'])
def industry_leader_maker_(cal_day, indicator, industry_weight, top_num,
index_for_simple_weight):
"""
:param stock_info: 当下A股市场的所有股票名称、代码信息
:param cal_day: 选股日期
:param indicator: 选股时引入的财务指标，包括总市值（total_mv）、流通市值（circ_mv）
:param industry_weight: 根据某指数，复制其行业权重。或者'market'：按全市场行业权重
:param top_num: 是否选择前几只
:param index_for_simple_weight:对于按某指数成分简单赋权
:return:
"""
# 读取股票信息表
stock_info = pd.read_excel(
'C:/Users/Lenovo/PycharmProjects/my_project/frjj_codes/stock_info_20190801.xlsx'
)
stock_info = stock_info.dropna(axis=0, how='any').reset_index(drop=True)
# 读取申万一级行业划分
sw1 = sw_info_(cal_day, "sw1")
sw1.rename(columns={'sw_name': 'sw1_name'}, inplace=True)
# 读取申万三级行业划分
sw3 = sw_info_(cal_day, "sw3")
sw3.rename(columns={'sw_name': 'sw3_name'}, inplace=True)
all = pd.DataFrame(columns=['code', 'code_name'],
index=range(len(stock_info)))
all['code'] = stock_info['证券代码']
all['code_name'] = stock_info['证券简称']
all = all.merge(sw1[['code', 'sw1_code', 'sw1_name']],
on='code',
how='left')
all = all.merge(sw3[['code', 'sw3_code', 'sw3_name']],
on='code',
how='left')
# 使cal_day适应api
cal_day = cal_day[0:4] + '-' + cal_day[4:6] + '-' + cal_day[6:8]
# 直接从数据库收集股票的流通市值（或总市值）
data = get_data_panel("huangwei",
"huangwei",
list(all['code']),
forwardchangeday2(cal_day),
forwardchangeday2(cal_day),
collection='stock_indicator')
# 替换原表的cir_mv
all = all.merge(data[['code', indicator]], on='code', how='left')
all['cir_mv'] = all[indicator]
del all[indicator]
# 计算申万三级行业内部流通市值比例
tot_industry = all.groupby('sw3_code').sum()
tot_industry['sw3_code'] = tot_industry.index
tot_industry = tot_industry.reset_index(drop=True)
all = all.merge(tot_industry, on='sw3_code', how='inner')
all['weight_in_industry'] = all['cir_mv_x'] / all['cir_mv_y']
all.rename(columns={
'cir_mv_x': 'cir_mv',
'cir_mv_y': 'industry_total'
},
inplace=True)
if industry_weight == 'market':
# 获得全市场每个申万一级行业的比重
tot_industry_sw1 = all.groupby('sw1_code').sum()
tot_industry_sw1['sw1_code'] = tot_industry_sw1.index
total_mv = tot_industry_sw1['cir_mv'].sum()
tot_industry_sw1[
'industryweight_in_all'] = tot_industry_sw1['cir_mv'] / total_mv
tot_industry_sw1.index.name = 'sw1_codes' # 更改level_name
all = all.merge(tot_industry_sw1[['industryweight_in_all',
'sw1_code']],
on='sw1_code',
how='inner')
else:
# 获得成分股中每个申万一级行业的比重
zz500_weight = get_data_crosssection_genaral("huangwei", "huangwei",
"index_weight",
'index_code',
industry_weight)
zz500_weight = zz500_weight[zz500_weight['date'] == list(
zz500_weight['date'][zz500_weight['date'] >= cal_day])
[0]].reset_index(drop=True)
# 与成分股权重表匹配最接近的日期
zz500_weight.rename(columns={'stock_code': 'code'}, inplace=True)
# 获取中证500中成分股的申万一级行业信息及流通市值信息
zz500_weight = zz500_weight.merge(all[['code', 'sw1_code', 'cir_mv']])
tot_industry_sw1 = zz500_weight.groupby('sw1_code').sum()
tot_industry_sw1['sw1_code'] = tot_industry_sw1.index
total_mv = tot_industry_sw1['cir_mv'].sum()
tot_industry_sw1[
'industryweight_in_all'] = tot_industry_sw1['cir_mv'] / total_mv
tot_industry_sw1.index.name = 'sw1_codes' # 更改level_name
all = all.merge(tot_industry_sw1[['industryweight_in_all',
'sw1_code']],
on='sw1_code',
how='inner')
# 添加申万三级行业总市值排名rank
tot_industry = tot_industry.sort_values(
by='cir_mv', ascending=False).reset_index(drop=True)
tot_industry['rank'] = tot_industry.index + 1
all = all.merge(tot_industry[['sw3_code', 'rank']],
on='sw3_code',
how='inner').sort_values(by='rank').reset_index(drop=True)
# 追加行业个股数量字段
count_industry = all.groupby('sw3_code').count()
count_industry['indu_num'] = count_industry['code']
count_industry['sw3_code'] = count_industry.index
count_industry.index.name = 'sw3_codes'
all = all.merge(count_industry[['sw3_code', 'indu_num']],
on='sw3_code',
how='inner')
# 每个行业先保留前5个
delete_list = []
indu_list = list(all['sw3_code'].unique())
for indu in indu_list:
one_indu = all[all['sw3_code'] == indu]
one_indu['global_index'] = one_indu.index
if one_indu['indu_num'].iloc[0] > 5:
one_indu = one_indu.sort_values(
by='weight_in_industry',
ascending=False).reset_index(drop=True)
delete_list = delete_list + list(one_indu['global_index'].iloc[5:])
top5_industry = all.drop(delete_list).reset_index(drop=True)
top5_industry['ten_percent'] = round(top5_industry['indu_num'] * 0.1)
delete_list = []
for indu in indu_list:
one_indu = top5_industry[top5_industry['sw3_code'] == indu]
one_indu['global_index'] = one_indu.index
if one_indu['ten_percent'].iloc[0] < 5:
one_indu = one_indu.sort_values(by='weight_in_industry',
ascending=False)
one_indu_withoutbig = one_indu[
one_indu['weight_in_industry'] < 0.2].reset_index(drop=True)
if len(one_indu_withoutbig) > 0:
remain_num = one_indu_withoutbig['ten_percent'].iloc[0] \
- (len(one_indu) - len(one_indu_withoutbig))
if remain_num <= 0:
delete_list = delete_list + list(
one_indu_withoutbig['global_index'])
else:
delete_list = delete_list + \
list(one_indu_withoutbig['global_index'].iloc[int(remain_num):])
result = top5_industry.drop(delete_list).reset_index(drop=True)
if top_num == None:
result_top100mv = result
else:
result_top100mv = result[result['rank'] <= top_num]
# 按成分股权重对龙头加权
hs300_weight = get_data_crosssection_genaral("huangwei", "huangwei",
"index_weight", 'index_code',
index_for_simple_weight)
hs300_weight = hs300_weight[hs300_weight['date'] == list(
hs300_weight['date'][hs300_weight['date'] >= cal_day])[0]].reset_index(
drop=True)
hs300_weight.rename(columns={'stock_code': 'code'}, inplace=True)
# 与沪深300股票取交集
result_top100mv = result_top100mv.merge(hs300_weight[['code', 'i_weight']],
on='code',
how='inner')
weight_sum = result_top100mv['i_weight'].sum()
result_top100mv['weight_adj'] = result_top100mv['i_weight'] / weight_sum
# 不与指数取交集，按照申万一级行业划分计算权重
tot_industry_sw1 = result.groupby('sw1_code').sum()
tot_industry_sw1['sw1_code'] = tot_industry_sw1.index
tot_industry_sw1 = tot_industry_sw1.reset_index(drop=True)
result = result.merge(tot_industry_sw1[['sw1_code', 'cir_mv']],
on='sw1_code',
how='inner')
result['weight_in_industry_sw1'] = result['cir_mv_x'] / result['cir_mv_y']
result.rename(columns={
'cir_mv_x': 'cir_mv',
'cir_mv_y': 'industry_total_sw1'
},
inplace=True)
# 申万一级行业的比重*行业内部个股占整个行业的比重
result['weight_adj_sw1'] = result['industryweight_in_all'] * result[
'weight_in_industry_sw1']
result['weight_adj_sw1'] = result['weight_adj_sw1'] / result[
'weight_adj_sw1'].sum()
# 还原cal_day
# cal_day = cal_day[0:4] + cal_day[5:7] + cal_day[8:10]
# file_name = 'result_' + cal_day + '.xls'
# result.set_index('code').to_excel(file_name, encoding='utf-8')
return result
for index, changeday in enumerate(change_days[:-2]):
# 读入股票代码
all_codes = pd.read_excel(
'C:/Users/Lenovo/PycharmProjects/my_project/frjj_codes/result/industry_leader_df_md_80.xlsx',
sheet_name=changeday)
part_codes = pd.read_excel(
'C:/Users/Lenovo/PycharmProjects/my_project/frjj_codes/result/industry_leader_df_md_50.xlsx',
sheet_name=changeday)
eighty_stock = set(all_codes['code'])
fifty_stock = set(part_codes['code'])
maybe_bad_stock = eighty_stock.difference(fifty_stock)
all_data_fifty_stock = get_data_panel("huangwei",
"huangwei",
list(fifty_stock),
changeday,
change_days[index + 1],
collection='stock_daily')
all_data_fifty_stock['rtn'] += 1
all_data_maybe_bad_stock = get_data_panel("huangwei",
"huangwei",
list(maybe_bad_stock),
changeday,
change_days[index + 1],
collection='stock_daily')
all_data_maybe_bad_stock['rtn'] += 1
# 分组计算累计收益率
fifty_stock_cum = all_data_fifty_stock.groupby(
by=all_data_fifty_stock['code']).apply(
lambda x: x['rtn'].cumprod().iloc[-1])
def factor_test(factor, types):
# 导入股票名单和相应权重
total_capitals = [10000000]
pnls = []
transaction_cost_rate = 0.003 # 交易费用
factor_list = [factor]
factor_type = [types]
result_export = False # 是否输出每天持仓信息
ic_weighter = True # 多因子是否按ic值加权平均
include_ir = True
orthogonalization = False
if result_export:
result_path = './result/' # 当下根目录
strategy_name = 'industry_leader'
file_name1 = strategy_name + '_df_md_80.xlsx'
# 获取所有调仓日
start, end = '20180101', "20190801"
change_days, my_tradingdays = get_time_data_(start, end)
# 开始回测
for changeday in tqdm(list(my_tradingdays['tradingdays'])):
if changeday == change_days[0]: # 建仓日
# stock_pool = GD_location_maker_(changeday)
stock_pool = leader_multifactor_maker(
changeday,
factor_list,
factor_type,
ic_weighter,
include_ir=include_ir,
orthogonalization=orthogonalization)
# stock_pool = leader_factor_maker(changeday)
# 持仓矩阵初始化
position = position_init_(stock_pool, changeday, total_capitals)
if result_export:
# 输出持仓信息
export_result_(position, changeday, result_path, file_name1)
# 记录当天的损益
pnls, total_capitals = return_maker_(position, pnls,
total_capitals,
transaction_cost_rate)
# last_stock_pool = stock_pool_class
# print(changeday, '开仓')
elif changeday != change_days[0] and changeday in change_days: # 调仓日
# 更新开盘价数据
price = get_data_panel("huangwei",
"huangwei",
list(position['code']),
changeday,
changeday,
collection='stock_daily').drop_duplicates(
['code'])
# 保存过去的持仓情况
last_position = position
if 'open' in last_position.columns:
del last_position['open']
if 'rtn' in last_position.columns:
del last_position['rtn']
last_position = last_position.merge(price[['code', 'open', 'rtn']],
on='code',
how='inner')
last_position['open_price'] = last_position['open']
last_position['close_price'] = last_position['open_price'] \
* (1 + last_position['rtn'])
# 按现在价格计算过去持仓的手数
last_position['lots'] = last_position['account'] / (
100 * last_position['open_price'])
last_position['lots'] = last_position['lots'].fillna(
0) # 没有开盘价的股票被“冻结”
last_position['lots'] = last_position['lots'].apply(
lambda x: math.floor(x))
# stock_pool = stock_select_(changeday)
# stock_pool, stock_pool_class = leader_moderate(changeday, last_stock_pool, 0.02)
stock_pool = leader_multifactor_maker(
changeday,
factor_list,
factor_type,
ic_weighter,
include_ir=include_ir,
orthogonalization=orthogonalization)
# stock_pool = leader_factor_maker(changeday)
# stock_pool = GD_location_maker_(changeday)
# 持仓初始化
position = position_init_(stock_pool, changeday, total_capitals)
position, adj_pnl = position_manager(changeday, last_position,
position,
transaction_cost_rate)
pnls.append(adj_pnl)
# 记录总价值
total_capitals.append(total_capitals[-1] + adj_pnl)
# 更新每只股票的总市值
position['account_realized'] = position['account_realized'] * (
1 + position['rtn'])
position['account'] = position['account_realized'] + position[
'account_surplus']
if result_export:
# 输出持仓信息
export_result_(position, changeday, result_path, file_name1)
# last_stock_pool = stock_pool_class
# print(changeday, '换仓')
else:
# 非调仓日
# 更新价格数据
position['date'] = changeday
price = get_data_panel("huangwei",
"huangwei",
list(position['code']),
changeday,
changeday,
collection='stock_daily').drop_duplicates(
['code'])
if 'open' in position.columns:
del position['open']
if 'rtn' in position.columns:
del position['rtn']
position = position.merge(price[['code', 'open', 'rtn']],
on='code',
how='left')
position['open_price'] = position['open']
position['close_price'] = position['open_price'] * (
1 + position['rtn'])
# 计算当日持仓（可能不是正确值，因为open价格未经调整，但可以作为一个基准计算合理的收益率）
position['account_realized'] = 100 * position[
'open_price'] * position['lots']
# 记录当天的损益
temp = position.dropna(axis=0)
pnl = (temp['close_price'] / temp['open_price'] -
1) * temp['account_realized']
pnls.append(pnl.sum())
# 记录总价值
total_capitals.append(total_capitals[-1] + pnl.sum())
# 更新每只股票的总市值
position['account_realized'] = position['account_realized'] * (
1 + position['rtn'])
position['account'] = position['account_realized'] + position[
'account_surplus']
# if result_export == True:
# # 输出持仓信息
# export_result_(position, changeday, result_path, file_name1)
# print(changeday, '持仓')
# 累计总价值曲线（与指数累计收益曲线作对比）
banchmark = "000300.SH"
returns = np.array(total_capitals)[1:] / np.array(total_capitals)[0]
hs300 = get_data_panel(
"huangwei",
"huangwei",
[banchmark], # 对标指数
my_tradingdays['tradingdays'].iloc[0],
my_tradingdays['tradingdays'].iloc[-1],
collection='index_rtn')
hs300['cum_chg'] = hs300['pct_chg'] / 100 + 1
hs300['cum_chg'] = hs300['cum_chg'].cumprod().values
extra = returns[:, np.newaxis] - hs300['cum_chg'].values[:, np.newaxis]
performance_(hs300, banchmark, returns, extra, factor_list[0], 'spyl')
return total_capitals
