def mfi_calc(self, df, print_log=True):
df['volume_aver'] = (df['volume_ask'] + df['volume_bid']) / 2
for tmprd_mfi in self.mfi_tmprd_arr:
postfix = '_' + processing.digit_to_text(tmprd_mfi)
new_clmn_names = [self.mfi_base_clmn_name + postfix]
if print_log: print('new_clmn_names: ', new_clmn_names)
inp = {
'df': df,
'function': tl.MFI,
'add_columns': new_clmn_names,
'shift': 1,
'high': df['high'].values,
'low': df['low'].values,
'close': df['close'].values,
'volume': df['volume_aver'].values,
'timeperiod': tmprd_mfi
}
df = processing.features_add(**inp)
for item in self.mfi_compare_arr:
clmn_1 = self.mfi_base_clmn_name + '_' + processing.digit_to_text(
item[0])
clmn_2 = self.mfi_base_clmn_name + '_' + processing.digit_to_text(
item[1])
new_clmn_name = self.mfi_base_clmn_name + '_cmpr_' + processing.digit_to_text(item[0]) + '_' + \
processing.digit_to_text(item[1])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
return df
def tema_calc(self, df, print_log=True):
for tmprd_tema in self.tema_tmprd_arr:
postfix = '_' + processing.digit_to_text(tmprd_tema)
new_clmn_names = [self.tema_base_clmn_name + postfix]
if print_log: print('new_clmn_names: ', new_clmn_names)
inp = {
'df': df,
'function': tl.TEMA,
'add_columns': new_clmn_names,
'shift': 0,
'real': df['open'].values,
'timeperiod': tmprd_tema
}
df = processing.features_add(**inp)
inp = (df, 'open', new_clmn_names[0], 'tema_open' + postfix)
df = processing.clmn_compare(*inp)
for item in self.tema_compare_arr:
clmn_1 = self.tema_base_clmn_name + '_' + processing.digit_to_text(
item[0])
clmn_2 = self.tema_base_clmn_name + '_' + processing.digit_to_text(
item[1])
new_clmn_name = self.tema_base_clmn_name + '_cmpr_' + processing.digit_to_text(item[0]) + '_' + \
processing.digit_to_text(item[1])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
return df
def run_adi_calc(self, df, dump_pickle=True, print_log=True):
time_start = dt.datetime.now()
if print_log: print('adi: time_start= {}'.format(time_start))
if print_log: print('cpu_count() = %d\n' % mp.cpu_count())
df['volume_aver'] = (df['volume_ask'] + df['volume_bid']) / 2
#--- Create pool
if print_log:
print('Creating pool with %d processes\n' % self.num_threads)
pool = mp.Pool(self.num_threads)
if print_log: print('pool = %s' % pool)
if print_log: print()
TASKS = []
for tmprd_adi in self.adi_tmprd_arr:
postfix = '_' + processing.digit_to_text(tmprd_adi)
new_clmn_names = [self.adi_base_clmn_name + postfix]
TASKS.append((df, new_clmn_names, tmprd_adi))
#print('TASKS:\n', TASKS)
results = [pool.apply_async(self.adi_calc, t) for t in TASKS]
for r in results:
res = r.get()
if print_log: print('\t', res)
if print_log:
print('type(res)= {0}, res= \n{1}'.format(type(res), res))
clmn_name = res[0][0]
if print_log: print('res[1][1]: \n', res[1][1])
clmn_data = res[1][1]
if print_log:
print('clmn_name= {0}, clmn_data:\n {1}'.format(
clmn_name, clmn_data))
df[clmn_name] = clmn_data
for item in self.adi_compare_arr:
clmn_1 = self.adi_base_clmn_name + '_' + processing.digit_to_text(
item[0])
clmn_2 = self.adi_base_clmn_name + '_' + processing.digit_to_text(
item[1])
new_clmn_name = self.adi_base_clmn_name + '_cmpr_' + processing.digit_to_text(item[0]) + '_' + \
processing.digit_to_text(item[1])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
if dump_pickle:
with open(self.data_pickle_path_for_dump, "wb") as pckl:
pickle.dump(df, pckl)
time_finish = dt.datetime.now()
time_duration = time_finish - time_start
if print_log:
print('adi: time_finish= {0}, duration= {1}'.format(
time_start, time_duration))
return df
def clmn_def(self):
postfix = '_' + processing.digit_to_text(self.trgt) + '_' + processing.digit_to_text(self.tpSL[0]) + \
'_' + processing.digit_to_text(self.tpSL[1])
self.__postfix = postfix
self.__buy_sl_clmn = 'buy_sl' + postfix
self.__buy_tp_clmn = 'buy_tp' + postfix
self.__sell_sl_clmn = 'sell_sl' + postfix
self.__sell_tp_clmn = 'sell_tp' + postfix
self.__labels_buy_clmn = self.lbl_bsc_names[0] + postfix
self.__labels_sell_clmn = self.lbl_bsc_names[1] + postfix
self.__target_clmn = self.lbl_bsc_names[2] + postfix
def macd_calc(self, df, print_log=True):
for params_macd in self.macd_params_arr:
postfix = '_' + processing.digit_to_text(params_macd[0]) \
+ '_' + processing.digit_to_text(params_macd[1]) + '_' + processing.digit_to_text(params_macd[2])
new_clmn_names = [
name + postfix for name in self.macd_base_clmn_names
]
if print_log: print('new_clmn_names: ', new_clmn_names)
inp = {
'df': df,
'function': tl.MACD,
'add_columns': new_clmn_names,
'shift': 0,
'real': df['open'].values,
'fastperiod': params_macd[0],
'slowperiod': params_macd[1],
'signalperiod': params_macd[2]
}
df = processing.features_add(**inp)
return df
def so_calc(self, df, print_log=True):
for param_so in self.so_param_arr:
postfix = '_' + processing.digit_to_text(
param_so[0]) + '_' + processing.digit_to_text(param_so[1])
new_clmn_names = [
name + postfix for name in self.so_base_clmn_names
]
if print_log: print('new_clmn_names: ', new_clmn_names)
inp = {
'df': df,
'function': tl.STOCH,
'add_columns': new_clmn_names,
'shift': 1,
'high': df['high'].values,
'low': df['low'].values,
'close': df['close'].values,
'fastk_period': param_so[0],
'slowk_period': int(param_so[0] * self.so_k_slowk_period),
'slowd_period': param_so[1]
}
df = processing.features_add(**inp)
inp = (df, new_clmn_names[0], new_clmn_names[1],
'so_k_d' + postfix)
df = processing.clmn_compare(*inp)
for item in self.so_compare_arr:
clmn_1 = 'so_k' + '_' + processing.digit_to_text(
item[0][0]) + '_' + processing.digit_to_text(item[0][1])
clmn_2 = 'so_k' + '_' + processing.digit_to_text(
item[1][0]) + '_' + processing.digit_to_text(item[1][1])
new_clmn_name = 'so_k_cmpr_' + processing.digit_to_text(item[0][0]) + '_' + \
processing.digit_to_text(item[0][1]) + '_' + processing.digit_to_text(item[1][0]) + '_' + \
processing.digit_to_text(item[1][1])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
return df
def return_calc(self, df, print_log=True):
for tmprd_rtrn in self.rtrn_tmprd_arr:
postfix = '_' + processing.digit_to_text(tmprd_rtrn)
new_clmn_names = [self.rtrn_base_clmn_name + postfix]
if print_log: print('new_clmn_names: ', new_clmn_names)
df[new_clmn_names[0]] = df['open'].rolling(
window=tmprd_rtrn,
center=False).apply(lambda x: self.return_feature(x))
for item in self.rtrn_compare_arr:
clmn_1 = self.rtrn_base_clmn_name + '_' + processing.digit_to_text(
item[0])
clmn_2 = self.rtrn_base_clmn_name + '_' + processing.digit_to_text(
item[1])
new_clmn_name = self.rtrn_base_clmn_name + '_cmpr_' + processing.digit_to_text(item[0]) + '_' + \
processing.digit_to_text(item[1])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
return df
def lr_calc(self, df, print_log=True):
for tmprd_lr in self.lr_tmprd_arr:
for mult_lr in self.lr_mult_arr:
postfix = '_' + processing.digit_to_text(
tmprd_lr) + '_' + processing.digit_to_text(mult_lr)
new_clmn_names = [
name + postfix for name in self.lr_base_clmn_names
]
if print_log: print('new_clmn_names: ', new_clmn_names)
tmprd_lr_1 = tmprd_lr
df[new_clmn_names[0]] = tl.LINEARREG_SLOPE(
df['open'].values, timeperiod=tmprd_lr_1)
tmprd_lr_2 = int(tmprd_lr_1 * mult_lr)
df[new_clmn_names[1]] = tl.LINEARREG_SLOPE(
df['open'].values, timeperiod=tmprd_lr_2)
inp = (df, new_clmn_names[0], new_clmn_names[1],
'lr_cmpr' + postfix)
df = processing.clmn_compare(*inp)
return df
def bb_calc(self, df, print_log=True):
for tmprd_bb in self.bb_tmprd_arr:
for d in self.bb_d_arr:
nbdevup, nbdevdn = d, d
postfix = '_' + processing.digit_to_text(
tmprd_bb) + '_' + processing.digit_to_text(d)
new_clmn_names = [
name + postfix for name in self.bb_base_clmn_names
]
if print_log: print('new_clmn_names: ', new_clmn_names)
inp = {
'df': df,
'function': tl.BBANDS,
'add_columns': new_clmn_names,
'real': df['open'].values,
'timeperiod': tmprd_bb,
'nbdevup': nbdevup,
'nbdevdn': nbdevdn,
'matype': 0
}
df = processing.features_add(**inp)
new_clmn_name = 'bb_rp' + postfix
df = self.bb_price_position(df, 'open', new_clmn_names[0],
new_clmn_names[1],
new_clmn_names[2], new_clmn_name)
# удаляем столбцы со значениями ubb, mbb, lbb. Остаётся только отн. положение цены в BB-канале.
df.drop(columns=new_clmn_names[0], inplace=True)
df.drop(columns=new_clmn_names[1], inplace=True)
df.drop(columns=new_clmn_names[2], inplace=True)
for item in self.bb_compare_arr:
clmn_1 = 'bb_rp' + '_' + processing.digit_to_text(
item[0]) + '_' + processing.digit_to_text(item[2])
clmn_2 = 'bb_rp' + '_' + processing.digit_to_text(
item[1]) + '_' + processing.digit_to_text(item[2])
new_clmn_name = 'bb_rp_cmpr_' + processing.digit_to_text(item[0]) + '_' + \
processing.digit_to_text(item[1]) + '_' + processing.digit_to_text(item[2])
df = processing.clmn_compare(df=df,
clmn_1=clmn_1,
clmn_2=clmn_2,
new_clmn_name=new_clmn_name)
return df
