import datetime
import requests
import json
import talib
import sys
sys.path.append("../")
from utils.plot import Plot
from service.calendar import Calendar
server_ip = "http://140.115.87.197:8090/"
cal = Calendar('TW')
payloads = {
'ticker_list': ['1524'],
'start_date': cal.get_trade_date('2010-01-01', (1+30)*-1, 'd'),
'end_date': cal.get_trade_date('2010-03-31', 1, 'd'),
}
response = requests.get(server_ip+"stk/get_ticker_period_stk", params=payloads)
stk_dict = json.loads(response.text)['result']
stk_df = pd.DataFrame(stk_dict[ticker_list[0]])
stk_df['date'] = [datetime.datetime.strptime(elm, "%Y-%m-%d") for elm in stk_df['date']]
stk_df.set_index("date", inplace=True)
stk_df.columns = ['Close', 'High', 'Low', 'Open', 'Volume', 'outstanding_share']
stk_df = stk_df.drop('outstanding_share', axis=1)
stk_df = stk_df.dropna()
print(stk_df)
up_band, mid, down_band = BBANDS(
stk_df['Close'], timeperiod=30,
class Analysis:
def __init__(self):
self._cfg = Config()
self._path = self._cfg.get_value('path', 'path_to_portfolio_performance')
self._api_server_IP = self._cfg.get_value('IP', 'api_server_IP')
self._start_equity = int(self._cfg.get_value('parameter', 'start_equity'))
self._start_date = self._cfg.get_value('parameter', 'start_date')
self._end_date = self._cfg.get_value('parameter', 'end_date')
self._general = General()
self._cal = Calendar('TW')
def plot_equity_curve(self, request):
perf_ind_inf_str = ""
equity_df, performance_df = self._read_request_files(request)
perf_ind, equity_max = self._compute_performance_ind(equity_df)
# 計算CAGR MDD MAR並製作顯示字串
for elm in perf_ind:
if elm != perf_ind[-1]:
perf_ind_inf_str = perf_ind_inf_str + "[{}] CAGR: {}, MDD: {}, MAR: {}\n".format(
elm[0], elm[1], elm[2], elm[3]
)
else:
perf_ind_inf_str = perf_ind_inf_str + "[{}] CAGR: {}, MDD: {}, MAR: {}".format(
elm[0], elm[1], elm[2], elm[3]
)
# 日期需設定為index
ax = equity_df.plot.line(figsize=(12, 8))
# ax = equity_df.plot.line(figsize=(12, 8), color=COLOR[0])
ax.set_xlabel('Date')
ax.set_ylabel('Equity')
# 加上千位數逗點
ax.get_yaxis().set_major_formatter(
FuncFormatter(lambda x, p: format(int(x), ','))
)
# 強調Y軸上某一個點
ax.axhline(10000000, color='black', linestyle='--')
# 顯示文字註解
plt.text(
# equity_df.index[200], equity_max/10*9,
equity_df.index[500], 8000000,
perf_ind_inf_str,
size=14, ha='left', va='top',
bbox=dict(
facecolor='gray', alpha=1.5, edgecolor='gray'
)
)
plt.show()
def plot_equity_curve_and_cap_util_rate(self, request):
perf_ind_inf_str = ""
equity_df, performance_df = self._read_request_files(request)
df = equity_df[[equity_df.columns[0]]].rename({equity_df.columns[0]: 0}, axis=1)
df.loc[:, :] = np.nan
position = performance_df['flow'].max()+1
for i in range(1, position):
df[i] = df[0]
for i, row in performance_df.iterrows():
df.loc[row['start']: row['end'], row['flow']] = 1
df['cap_util'] = df.iloc[:, :].sum(axis=1)
df['cap_util_rate'] = df['cap_util'].apply(lambda x: x/position*100)
df['cap_util_rate'].iloc[-1] = np.nan
equity_df = equity_df.join(df['cap_util_rate'])
perf_ind, equity_max = self._compute_performance_ind(equity_df.iloc[:, :-1])
# 計算CAGR MDD MAR並製作顯示字串
for elm in perf_ind:
if elm != perf_ind[-1]:
perf_ind_inf_str = perf_ind_inf_str + "[{}] CAGR: {}%, MDD: {}%\n".format(
elm[0], elm[1], elm[2]
)
else:
perf_ind_inf_str = perf_ind_inf_str + "[{}] CAGR: {}%, MDD: {}%".format(
elm[0], elm[1], elm[2]
)
perf_ind_inf_str = perf_ind_inf_str + "\nFull Use of Equity Rate: {}%".format(
100 - round(df.loc[df['cap_util'] != position].shape[0] / df.shape[0] * 100, 2)
)
# ax = equity_df.iloc[:, :-1].plot.line(figsize=(12, 8))
ax = equity_df.iloc[:, :-1].plot.line(figsize=(12, 8), color=[COLOR[1], COLOR[2]])
# 顯示文字註解
plt.text(
# equity_df.index[1], equity_df[equity_df.columns[0]].min(),
equity_df.index[0], 3580000,
perf_ind_inf_str,
size=13, ha='left', va='top',
bbox=dict(
facecolor='gray', alpha=1.5, edgecolor='gray'
)
)
ax2 = equity_df['cap_util_rate'].plot.line(color='grey', secondary_y=True)
# 強調Y軸上某一個點
ax.axhline(10000000, color='black', linestyle='--')
# 加上千位數逗點
ax.get_yaxis().set_major_formatter(
FuncFormatter(lambda x, p: format(int(x), ','))
)
ax.set_xlabel("Date")
ax.set_ylabel("Equity")
ax2.set_ylabel("Capital Utilization Rate[%]")
plt.ylim(-10, 110)
plt.show()
def plot_performance_heatmap(self, request):
heatmap_df = pd.DataFrame(columns=[6, 15, 60, 150, 300])
df = self.read_performance_file()
df = df.loc[
(df['factor'] == request['factor']) &
(df['strategy'] == request['strategy']) &
(df['window'] == request['window']) &
(df['method'] == request['method'])
]
for i in [1, 2, 3, 4, 5]:
cagr_list = []
for col in heatmap_df.columns:
cagr_list.append(
df[request['perf_ind']].loc[
(df['group']==i) & (df['position']==col)
].values[0]
)
heatmap_df.loc[i] = cagr_list
print(heatmap_df)
ax = sns.heatmap(heatmap_df, cmap="YlGnBu")
ax.set_xlabel("Position")
ax.set_ylabel("Group")
# ax.set_title("[{}] {}_{}_{}".format(
# request['perf_ind'], request['factor'],
# request['window'], request['method'])
# )
print("[{}] {}_{}_{}_{}".format(
request['perf_ind'], request['factor'],
request['strategy'], request['window'],
request['method']
)
)
plt.show()
def plot_profit_bar_chart(self, request):
equity_df, trade_df = self._read_request_files(request)
file_name = equity_df.columns[0]
trade_df['win'] = trade_df['return'].apply(lambda x: 1 if x > 0 else 0)
win_rate = round(trade_df['win'].sum() / trade_df.shape[0] * 100, 2)
lose_rate = round(100 - win_rate, 2)
trade_df['+0-20'] = trade_df['return'].apply(lambda x: 1 if x > 0 and x <= 20 else 0)
trade_df['+20-100'] = trade_df['return'].apply(lambda x: 1 if x > 20 and x <= 100 else 0)
trade_df['+100<'] = trade_df['return'].apply(lambda x: 1 if x > 100 else 0)
trade_df['-0-20'] = trade_df['return'].apply(lambda x: 1 if x <= 0 and x >= -20 else 0)
trade_df['-20>'] = trade_df['return'].apply(lambda x: 1 if x < -20 else 0)
positive_df = trade_df.copy()
positive_df['return'] = positive_df['return'].apply(lambda x: 0 if x < 0 else x)
negative_df = trade_df.copy()
negative_df['return'] = negative_df['return'].apply(lambda x: 0 if x >= 0 else x)
text_str = "{}\n--------------------------------\n100 <: {}\n20 ~ 100: {}\n0 ~ 20: {}\n--------------------------------\n0 ~ -20: {}\n-20 >: {}\n--------------------------------\nTotal Trades Number: {}\nWin Rate: {}\nLose Rate: {}".format(
file_name,
trade_df['+100<'].sum(), trade_df['+20-100'].sum(), trade_df['+0-20'].sum(),
trade_df['-0-20'].sum(), trade_df['-20>'].sum(),
len(trade_df.index), win_rate, lose_rate
)
fig = plt.figure(figsize=(80, 8))
sns.set_theme(style="whitegrid")
ax = sns.barplot(x="index", y="return", data=positive_df.reset_index(), color="navy")
ax = sns.barplot(x="index", y="return", data=negative_df.reset_index(), color="gray")
plt.legend()
plt.xlabel("Trade Series Number")
plt.ylabel("Return[%]")
plt.xticks(fontsize=2, rotation=90)
plt.yticks(fontsize=12)
# 顯示文字註解
plt.text(
trade_df.index[200], trade_df['return'].max(),
# trade_df.index[180], 90,
text_str,
size=14, ha='left', va='top',
bbox=dict(
facecolor='gray', alpha=1.5, edgecolor='gray'
)
)
plt.show()
def plot_window_profit_bar_chart(self, request):
equity_df, _ = self._read_request_files(request)
file_name = equity_df.columns[0]
report_date_list = self._cal.get_report_date_list(self._start_date, self._end_date)
window_return = []
peroid_list = []
# 窗格數量會比期間內財報公布日數量多一個
for i in range(len(report_date_list)+1):
# 第一個窗格T2: 回測期間第一個交易日~第一個財報公布日
if i == 0:
date = self._cal.get_trade_date(self._start_date, 0, 'd')
peroid_list.append(date)
strat_equity = equity_df.loc[date, file_name]
final_equity = equity_df.loc[report_date_list[i], file_name]
final_equity = final_equity.iloc[0]
# 最後一個窗格T2: 最後一個財報公布日~回測期間最後一個交易日
elif i == len(report_date_list):
date = self._cal.get_trade_date(self._end_date, -1, 'd')
peroid_list.append(report_date_list[i-1])
strat_equity = equity_df.loc[report_date_list[i-1], file_name]
final_equity = equity_df.loc[date, file_name]
strat_equity = strat_equity.iloc[0]
# 其他窗格T2: 兩公布日之間
else:
peroid_list.append(report_date_list[i-1])
strat_equity = equity_df.loc[report_date_list[i-1], file_name]
final_equity = equity_df.loc[report_date_list[i], file_name]
strat_equity = strat_equity.iloc[0]
final_equity = final_equity.iloc[0]
window_return.append(round((final_equity-strat_equity)/strat_equity*100, 2))
window_return_df = pd.DataFrame(window_return, columns=['return'])
window_return_df['window_peroid'] = peroid_list
total_return = (equity_df.iloc[-1, 0]-10000000) / 10000000 * 100
positive_df = window_return_df.copy()
positive_df['return'] = positive_df['return'].apply(lambda x: 0 if x < 0 else x)
negative_df = window_return_df.copy()
negative_df['return'] = negative_df['return'].apply(lambda x: 0 if x >= 0 else x)
result, _ = self._compute_performance_ind(equity_df)
text_str = "{}\nCAGR: {}%, MDD: {}%".format(
file_name, result[0][1], result[0][2]
)
fig = plt.figure(figsize=(12, 7))
sns.set_theme(style="whitegrid")
ax = sns.barplot(x="window_peroid", y="return", data=positive_df, color="cornflowerblue")
ax = sns.barplot(x="window_peroid", y="return", data=negative_df, color="gray")
plt.legend()
plt.xlabel("Start Date of Window", fontsize=8)
plt.ylabel("Return[%]")
plt.xticks(fontsize=8, rotation=35)
plt.yticks(fontsize=12)
# 顯示文字註解
plt.text(
17, window_return_df['return'].max()-2,
text_str,
size=13, ha='left', va='top',
bbox=dict(
facecolor='gray', alpha=1.5, edgecolor='gray'
)
)
plt.show()
def plot_linear_regression(self, request):
p_df = self.read_performance_file()
lr_df = self.read_regression_file()
factor_str_list = []
query_index = "{}_{}_{}_{}".format(
request['strategy'], request['window'],
request['method'], request['position'],
)
print(lr_df.loc[
(lr_df['perf_ind'] == request['perf_ind']) &
(lr_df['index'] == query_index)
])
lr_df = lr_df.loc[
(lr_df['perf_ind'] == request['perf_ind']) &
(lr_df['index'] == query_index)
].set_index('index').T.drop('perf_ind')
print(lr_df.sort_values(by=[query_index]))
path = self._cfg.get_value('path', 'path_to_performance_analysis')
lr_df.sort_values(by=[query_index]).to_csv(path+'lr_df.csv')
fig = plt.figure(figsize=(8, 7))
for i, factor in enumerate(request['factor_list']):
factor_str = self._general.factor_to_string(factor)
factor_str_list.append(factor_str)
temp_df = p_df.loc[
(p_df['factor'] == factor_str) &
(p_df['strategy'] == request['strategy']) &
(p_df['window'] == request['window']) &
(p_df['method'] == request['method']) &
(p_df['position'] == request['position'])
]
ax = sns.regplot(x=temp_df['group'], y=temp_df[request['perf_ind']], color=COLOR[i])
ax.set(xticks=range(1, len(temp_df['group'])+1, 1))
plt.legend(labels=factor_str_list)
plt.xlabel('Group', fontsize=12)
plt.ylabel(request['perf_ind'], fontsize=12)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12)
plt.show()
def output_performance_file(self, request):
performance_df = pd.DataFrame(
columns=[
'factor', 'strategy', 'window', 'method', 'group',
'position', 'CAGR[%]', 'MDD[%]', 'MAR'
]
)
response = requests.get("http://{}/task/get_task_combi".format(self._api_server_IP), params=request)
combination = json.loads(response.text)['result']
for combi in combination:
equity_df, _ = self._read_file(combi)
ind_list, _ = self._compute_performance_ind(equity_df)
performance_df = performance_df.append({
'factor': self._general.factor_to_string(combi[0]),
'strategy': combi[1],
'window': combi[2],
'method': combi[3],
'group': combi[4],
'position': combi[5],
'CAGR[%]': ind_list[0][1],
'MDD[%]': ind_list[0][2],
'MAR': ind_list[0][3],
}, ignore_index=True)
print(performance_df)
path = self._cfg.get_value('path', 'path_to_performance_analysis')
performance_df.to_csv(path+'performance_indicator.csv')
# performance_df.to_csv(path+'performance_indicator_a&na.csv')
def output_linear_regression_file(self, request):
df = self.read_performance_file()
combination = []
lr_df = pd.DataFrame()
column_order = ['index', 'perf_ind']
data = (
request['strategy_list'], request['window_list'],
request['method_list'], request['position_list']
)
for combi in product(*data):
# strategy: B&H
if combi[0] == 0:
# B&H 排除0之外的窗格 排除等權重之外的方法
if combi[1] != 0 or combi[2] != 0:
continue
combination.append(combi)
for perf_ind in request['perf_ind_list']:
for c in combination:
factor_b1_val_dict = {}
factor_b1_val_dict["index"] = "{}_{}_{}_{}".format(
c[0], c[1], c[2], c[3]
)
for i, factor in enumerate(request['factor_list']):
factor_str = self._general.factor_to_string(factor)
# if len(column_order) != 14:
if len(column_order) != 3:
column_order.append(factor_str)
temp_df = df.loc[
(df['factor'] == factor_str) &
(df['strategy'] == c[0]) &
(df['window'] == c[1]) &
(df['method'] == c[2]) &
(df['position'] == c[3])
]
x = self._compute_simple_linear_regression(temp_df['group'], temp_df[perf_ind])
factor_b1_val_dict[factor_str] = round(x[1], 3)
factor_b1_val_dict['perf_ind'] = perf_ind
lr_df = lr_df.append(factor_b1_val_dict, ignore_index=True)
lr_df = lr_df[column_order]
print(lr_df)
path = self._cfg.get_value('path', 'path_to_performance_analysis')
lr_df.to_csv(path+'regression_b1.csv')
# lr_df.to_csv(path+'regression_b1_a&na.csv')
def read_performance_file(self):
path = self._cfg.get_value('path', 'path_to_performance_analysis')
df = pd.read_csv(path+'performance_indicator.csv').drop('Unnamed: 0', axis=1)
# df = pd.read_csv(path+'performance_indicator_a&na.csv').drop('Unnamed: 0', axis=1)
return df
def read_regression_file(self):
path = self._cfg.get_value('path', 'path_to_performance_analysis')
df = pd.read_csv(path+'regression_b1.csv').drop('Unnamed: 0', axis=1)
# df = pd.read_csv(path+'regression_b1_a&na.csv').drop('Unnamed: 0', axis=1)
return df
def _compute_simple_linear_regression(self, raw_x, raw_y):
n = np.size(raw_x)
x = np.array(raw_x)
y = np.array(raw_y)
x_mean = np.mean(x)
y_mean = np.mean(y)
num1 = np.sum(y*x) - n*y_mean*x_mean
num2 = np.sum(x*x) - n*x_mean*x_mean
b_1 = num1 / num2
b_0 = y_mean - b_1 * x_mean
return (b_0, b_1)
def _compute_performance_ind(self, df):
result = []
equity_max = 0
for col in df.columns:
equity_df = df[col]
# [CAGR]
start_date_list = self._start_date.split("-")
end_date_list = self._end_date.split("-")
final_equity = equity_df.iloc[-1]
if end_date_list[1] == "1" and end_date_list[2] == "1":
n = int(end_date_list[0]) - int(start_date_list[0])
else:
n = int(end_date_list[0]) - int(start_date_list[0]) + 1
cagr = ((final_equity / self._start_equity) ** (1/n) - 1) * 100
# [MDD]
highest_equity = equity_df.max()
MDD = 0
for i, elm in enumerate(equity_df.values.tolist()):
drawdown = elm - equity_df.iloc[:i].max()
if drawdown < MDD:
MDD = drawdown
MDD = MDD / highest_equity *100
MAR = cagr / abs(MDD)
result.append([col, round(cagr, 2), round(MDD, 2), round(MAR, 3)])
if equity_df.max() > equity_max:
equity_max = equity_df.max()
return result, equity_max
def _read_file(self, param):
# 將[fac1, fac2] 轉成 fac1&fac2
factor_str = self._general.factor_to_string(param[0])
path = "{}{}/".format(self._path, factor_str)
file_name = "{}_{}_{}_{}_{}_{}".format(
factor_str,
param[1], # strategy
param[2], # window
param[3], # method
param[4], # group
param[5], # position
)
performance_df = pd.read_csv("{}{}.csv".format(path, file_name)).drop('Unnamed: 0', axis=1)
with open(path+file_name+'.json', 'r') as file:
json_data = json.load(file)
equity_df = pd.DataFrame.from_dict(json_data)
equity_df['date']= pd.to_datetime(equity_df['date'])
equity_df = equity_df.set_index('date')
equity_df.columns = [file_name]
print("read {}".format(file_name))
return equity_df, performance_df
def _read_request_files(self, request):
equity_df = pd.DataFrame()
response = requests.get("http://{}/task/get_task_combi".format(self._api_server_IP), params=request)
combination = json.loads(response.text)['result']
# combination = [
# [['EV_EBITDA'], 0, 0, 0, 1, 6],
# [['EV_EBITDA'], 1, 0, 0, 1, 6],
# [['EV_EBITDA'], 1, 1, 0, 1, 6],
# [['EV_EBITDA'], 1, 2, 0, 1, 6],
# ]
# 讀檔並串接 equity
for param in combination:
temp_equity_df, performance_df = self._read_file(param)
if equity_df.empty:
equity_df = temp_equity_df
else:
equity_df = equity_df.join(temp_equity_df)
return equity_df, performance_df