def back_test(self,horizon,istech,large,freq,model_name,address,roll=-1,threshold=0.1):
# initial setting
df = self.context_dict['close'].copy()
symbols = self.context_dict['close'].columns[1:]
cols = self.context_dict['close'].columns.values
cols[0]='ben'
df.columns = cols# set zz500 as benchmark
stock_num = len(symbols)
back_testing = df.index[pd.to_datetime(df.index) >= pd.to_datetime(self.start_day)]
df = df.loc[back_testing.values, :]
unit = np.full((len(df.index), 1), 1)[:, 0]
df['rebalancing'] = pd.Series()
df['stoploss'] = pd.Series()
df['nav'] = pd.Series(unit, index=df.index)
df['Interest_rate'] = pd.Series(np.full((len(df.index),), 2.5),index=df.index) # 2.5% interest_rate
weight_new = []
# max_new = [] # for computing max_drawdown
unit = np.full((len(df.index), stock_num), 0)
weights = pd.DataFrame(unit, index=df.index, columns=symbols)
reb_index = 0
s = 0 # counting date
# ============================= Enter Back-testing ===================================
for cur_date in back_testing.values:
cur_date = pd.to_datetime(cur_date)
# rebalance in a fixed frequency in freq rate
if s>0:# begin to rebalance at least after the second recordings
if np.mod(s, freq) == 0:
# print(s)
if self.extract_train(cur_date, istech,horizon,large =large,roll=roll,):
if np.shape(self.x_test)[0]>0:
test_y = strats.model(model_name, self.x_train, self.y_train, self.x_test)
weight_new = strats.fix_stock_order(test_y, threshold)
print(df.index[s - 1])
print(weight_new[weight_new != 0].head())
weights.loc[df.index[s - 1], weight_new.index] = weight_new.values
print('*' * 50)
df['rebalancing'].iloc[s - 1] = 1
reb_index = s - 1
if len(weight_new) != 0:
df = ut.record_return(df, s, reb_index, weight_new, self.leverage, self.trading_days)
weights = ut.record_weights(df, s, weights)
s += 1 # counting date
compute_indicators(df, 'perf_'+address,self.trading_days)
weights.to_csv('weights_' + address)
print('back_test completed!')
max_new = pd.Series(temp_max, index=df.columns[:stock_num])
# rebalance in fixed frequency
if np.mod(i - return_period, freq) == 0:
# The price data of etf uis from the second df.column
weight_new = ut.mk_drawdown(df.iloc[:i, :stock_num], back_max,
back_holding_period, return_period,
cov_period)
# weight_new = fix_fund_order(weight_new)
# TODO:how to record weights?
print(df.index[i - 1])
print(weight_new)
print('*' * 50)
df['rebalancing'].ix[i - 1] = 1
reb_index = i - 1
# stoploss
if len(max_new) != 0:
weight_new, flag = stoploss(df, stop_loss, i, max_new, weight_new)
if flag == 1:
print(df.index[i - 1])
print('stoploss!: ')
print(weight_new)
print('-' * 50)
df['stoploss'].ix[i - 1] = 1
reb_index = i - 1
if len(weight_new) != 0:
df = ut.record_return(df, stock_num, i, reb_index, weight_new,
leverage, trading_days)
perf = ut.comput_idicators(df, trading_days, required_return_real,
save_file, save_address + '.csv')