def start(date_now, result):
if Date().is_workday(date_now):
for ts_code in ['399300.SZ']:
try:
start_date, end_date = date_now - datetime.timedelta(days=30), date_now
security_point_data = SecurityData().get_index_point_data(ts_code, start_date, end_date)
buy_flag = calc_bs_data(security_point_data)
if buy_flag is True:
previous_work_day = Date().get_previous_workday(end_date)
previous_2_work_day = Date().get_previous_workday(previous_work_day)
next_0_workday = date_now
next_1_workday = Date().get_next_workday(date_now)
next_2_workday = Date().get_next_workday(next_1_workday)
next_3_workday = Date().get_next_workday(next_2_workday)
next_4_workday = Date().get_next_workday(next_3_workday)
next_5_workday = Date().get_next_workday(next_4_workday)
security_point_data = SecurityData().get_qfq_security_point_data(ts_code, previous_2_work_day,
end_date + datetime.timedelta(days=20))
security_data = security_point_data
result_now = pd.DataFrame(
[security_data.loc[previous_2_work_day], security_data.loc[previous_work_day], security_data.loc[next_0_workday],
security_data.loc[next_1_workday], security_data.loc[next_2_workday], security_data.loc[next_3_workday],
security_data.loc[next_4_workday], security_data.loc[next_5_workday]],
index=[['previous_2_day', 'previous_1_day', 'next_0_day', 'next_1_day', 'next_2_day', 'next_3_day', 'next_4_day',
'next_5_day'], [1, 1, 1, 1, 1, 1, 1, 1]])
result_now = result_now.unstack(level=0)
result.append(result_now)
except Exception as e:
pass
return result
def start(date_now=None):
date_now = datetime.datetime.now() if date_now is None else date_now
date_now = datetime.datetime(date_now.year, date_now.month, date_now.day)
if Date().is_workday(date_now):
start_date, end_date = date_now - datetime.timedelta(
days=365), date_now
ts_codes = MarketInfo().get_future_main_code_filter_by_manual(date_now)
for ts_code in ts_codes:
try:
security_point_data = SecurityData(
).get_future_security_point_data(ts_code, start_date, end_date)
buy_flag = buy(security_point_data)
if buy_flag is True:
Result().insert_strategy_result_data(
ts_code, "future_bs_when_trend_start", "B", date_now)
except Exception as e:
pass
try:
security_point_data = SecurityData(
).get_future_security_point_data(ts_code, start_date, end_date)
buy_flag = sell(security_point_data)
if buy_flag is True:
Result().insert_strategy_result_data(
ts_code, "future_bs_when_trend_start", "S", date_now)
except Exception as e:
pass
def start(date_now=None):
date_now = datetime.datetime.now() if date_now is None else date_now
date_now = datetime.datetime(date_now.year, date_now.month, date_now.day)
if Date().is_workday(date_now):
start_date, end_date = date_now - datetime.timedelta(days=50), date_now
# ts_codes = Security().get_efficient_ts_code_by_column_name_list(["normal_status", "tactics_5_status"])
# for ts_code in ts_codes:
for ts_code in [
'600481.SH', '601699.SH', '002061.SZ', '002424.SZ',
'600436.SH', '002796.SZ', '600810.SH', '000510.SZ',
'601600.SH', '600218.SH', '002414.SZ', '002302.SZ',
'000877.SZ', '002451.SZ', '600355.SH', '600760.SH',
'600679.SH', '000656.SZ', '600148.SH', '000935.SZ',
'002208.SZ', '600516.SH', '002602.SZ', '002696.SZ',
'002050.SZ', '000557.SZ', '002460.SZ', '600230.SH',
'002877.SZ', '002243.SZ', '600235.SH', '000403.SZ',
'000576.SZ', '600997.SH', '600581.SH', '600690.SH',
'600507.SH', '600019.SH', '000513.SZ', '002346.SZ',
'600966.SH', '000612.SZ', '000966.SZ', '600438.SH',
'002382.SZ', '600388.SH', '603993.SH', '601636.SH',
'000898.SZ', '601028.SH', '000739.SZ', '002258.SZ',
'002760.SZ', '603288.SH', '600782.SH', '600720.SH',
'002708.SZ', '603027.SH', '000785.SZ', '600141.SH',
'000001.SZ', '600783.SH', '600519.SH', '002072.SZ',
'600280.SH', '600874.SH', '603859.SH', '601003.SH',
'601918.SH', '000732.SZ', '600083.SH', '601933.SH',
'600425.SH', '002034.SZ', '600346.SH', '000528.SZ',
'601360.SH', '002679.SZ', '601318.SH', '002124.SZ',
'603444.SH', '002606.SZ', '002157.SZ', '600596.SH',
'600761.SH', '002408.SZ', '600815.SH', '603180.SH',
'000725.SZ', '600641.SH', '000830.SZ', '600128.SH',
'002056.SZ', '600196.SH', '600104.SH', '600080.SH',
'002545.SZ', '601012.SH', '600776.SH', '000717.SZ',
'000690.SZ', '600053.SH', '601100.SH', '002415.SZ'
]:
try:
security_point_data = SecurityData().get_security_point_data(
ts_code, start_date, end_date)
buy_flag = buy(security_point_data)
if buy_flag is True:
Result().insert_strategy_result_data(
ts_code, "sma_sloop", "B", date_now)
except Exception as e:
pass
ts_codes = Result().get_hold_data('sma_sloop')
for ts_code in ts_codes:
try:
security_point_data = SecurityData().get_security_point_data(
ts_code, start_date, end_date)
sell_flag = sell(security_point_data)
if sell_flag is True:
Result().insert_strategy_result_data(
ts_code, "sma_sloop", "S", date_now)
except Exception as e:
pass
def start():
security_data = SecurityData().get_index_point_data(
'399300.SZ', datetime.datetime(2016, 5, 3), datetime.datetime.now())
hs300(security_data)
security_data = SecurityData().get_security_point_data(
'600030.SH', datetime.datetime(2016, 5, 3), datetime.datetime.now())
zx_security(security_data)
def start(all_ts_code, date):
"""失败 准确率太低"""
for ts_code in all_ts_code:
print(ts_code)
qfq_security_point_data = SecurityData().get_qfq_security_point_data(ts_code, datetime.datetime(2016, 5, 4), datetime.datetime(2019, 1, 31))
qfq_security_point_data.sort_index(ascending=False, inplace=True)
qfq_security_point_data.index.rename("date", inplace=True)
security_daily_basic = SecurityData().get_security_daily_basic_data(ts_code, datetime.datetime(2016, 5, 4), datetime.datetime(2019, 1, 31))
security_daily_basic.sort_index(ascending=False, inplace=True)
security_daily_basic.index.rename("date", inplace=True)
target_name_list = ["next_2_day", "next_3_day", "next_4_day", "next_5_day", "next_6_day", "next_7_day", "next_8_day", "next_9_day", "next_10_day"]
target_data_list = []
for window_num in range(3, 12):
target_data_now = qfq_security_point_data["close"].rolling(window=window_num).apply(lambda x: x[0] > x[-1], raw=True)
target_data_now = target_data_now.iloc[12:]
target_data_list.append(target_data_now.values)
percent_series = qfq_security_point_data["open"] / 100
data = pd.DataFrame({
"open": qfq_security_point_data["open"] / percent_series,
"close": qfq_security_point_data["close"] / percent_series,
"high": qfq_security_point_data["high"] / percent_series,
"low": qfq_security_point_data["low"] / percent_series,
"turnover_rate_f": security_daily_basic["turnover_rate_f"]
}, index=qfq_security_point_data["trade_date"])
data = data.iloc[12:]
data = data.fillna(0)
data = data.values
for day, target_data in zip(target_name_list, target_data_list):
print("day is {0}".format(day))
depth_list = []
training_score_list = []
test_score_list = []
for max_depth in range(1, 31):
X_train, X_test, y_train, y_test = train_test_split(data, target_data, random_state=0)
forest = RandomForestClassifier(n_estimators=100, random_state=0, max_depth=max_depth)
forest.fit(X_train, y_train)
print("Accuracy on training set: {:.3f}".format(forest.score(X_train, y_train)))
print("Accuracy on test set: {:.3f}".format(forest.score(X_test, y_test)))
if forest.score(X_test, y_test) > 0.7:
training_score_list.append(forest.score(X_train, y_train))
test_score_list.append(forest.score(X_test, y_test))
depth_list.append(max_depth)
if test_score_list:
# plt.figure(figsize=(22, 15))
plt.figure(figsize=(6, 6))
plt.locator_params(nbins=30, axis='x')
plt.plot(depth_list, training_score_list, ls="-", lw=2, label="training figure")
plt.plot(depth_list, test_score_list, ls="-", lw=2, label="test figure") # ls: 折线图线条风格, lw: 线条宽度, label: 标签文本
plt.title(day)
plt.legend() # 用以显示label的内容
# plt.show()
plt.savefig('./' + ts_code + 'day_' + str(day) + '.png')
print("\n\n\n\n\n")
aa = 1
def start(date_now, result):
if Date().is_workday(date_now):
ts_codes = Security().get_efficient_ts_code_by_column_name_list(
["normal_status", "tactics_5_status"])
# ts_codes = ['002062.SZ', '002565.SZ']
for ts_code in ts_codes:
try:
start_date, end_date = date_now - datetime.timedelta(
days=30), date_now
security_point_data = SecurityData(
).get_qfq_security_point_data(ts_code, start_date, end_date)
buy_flag = calc_bs_data(security_point_data)
if buy_flag is True:
previous_work_day = Date().get_previous_workday(end_date)
previous_2_work_day = Date().get_previous_workday(
previous_work_day)
security_point_data = SecurityData(
).get_qfq_security_point_data(
ts_code, previous_2_work_day,
end_date + datetime.timedelta(days=20))
result_now = pd.DataFrame([
security_point_data.loc[previous_2_work_day],
security_point_data.loc[previous_work_day],
security_point_data.iloc[2],
security_point_data.iloc[3],
security_point_data.iloc[4],
security_point_data.iloc[5],
security_point_data.iloc[6],
security_point_data.iloc[7]
],
index=[[
'previous_2_day',
'previous_1_day',
'next_0_day', 'next_1_day',
'next_2_day', 'next_3_day',
'next_4_day', 'next_5_day'
], [1, 1, 1, 1, 1, 1, 1, 1]])
result_now = result_now.unstack(level=0)
result.append(result_now)
except Exception as e:
pass
return result
def start(date_now=None):
date_now = datetime.datetime.now() if date_now is None else date_now
date_now = datetime.datetime(date_now.year, date_now.month, date_now.day)
if Date().is_workday(date_now):
start_date, end_date = date_now - datetime.timedelta(days=50), date_now
for ts_code in ['600406.SH', '600298.SH', '000001.SZ', '600346.SH', '600104.SH', '601088.SH', '600028.SH', '600340.SH', '600031.SH', '600967.SH',
'002216.SZ', '600498.SH', '600176.SH', '002487.SZ', '600857.SH', '603959.SH', '002242.SZ', '603160.SH']:
try:
security_point_data = SecurityData().get_security_point_data(ts_code, start_date, end_date)
buy_flag = buy(security_point_data)
if buy_flag is True:
Result().insert_strategy_result_data(ts_code, "buy_when_fall", "B", date_now)
except Exception as e:
pass
def start(date_now=None):
date_now = datetime.datetime.now() if date_now is None else date_now
date_now = datetime.datetime(date_now.year, date_now.month, date_now.day)
if Date().is_workday(date_now):
ts_codes = Security().get_efficient_ts_code_by_column_name_list(
["normal_status", "tactics_5_status"])
for ts_code in ts_codes:
try:
start_date, end_date = date_now - datetime.timedelta(
days=90), date_now
security_point_data = SecurityData().get_security_point_data(
ts_code, start_date, end_date)
buy_flag = calc_bs_data(security_point_data)
if buy_flag is True:
Result().insert_strategy_result_data(
ts_code, "bbond", "B", date_now)
except Exception as e:
pass
def start(date_now=None):
date_now = datetime.datetime.now() if date_now is None else date_now
date_now = datetime.datetime(date_now.year, date_now.month, date_now.day)
if Date().is_workday(date_now):
ts_codes = Security().get_efficient_ts_code_by_column_name_list(
["normal_status", "tactics_5_status", "tactics_7_status"])
start_date, end_date = date_now - datetime.timedelta(
days=465), date_now
for ts_code in ts_codes:
# for ts_code in '000592.SZ', '603800.SH', '603700.SH', '002547.SZ', '000709.SZ', '002006.SZ', '002375.SZ', '002009.SZ', '002873.SZ', '002883.SZ', '002911.SZ', '002862.SZ', '601005.SH':
try:
security_point_data = SecurityData().get_security_point_data(
ts_code, start_date, end_date)
holder_number_data = MarketData().get_holder_number_data(
ts_code, start_date, end_date)
buy_flag = buy(security_point_data, holder_number_data)
if buy_flag is True:
Result().insert_strategy_result_data(
ts_code, "buy_when_holder_number_fall", "B", date_now)
except Exception as e:
pass
def start(date_now, result):
if Date().is_workday(date_now):
ts_codes = Security().get_efficient_ts_code_by_column_name_list(
["normal_status", "tactics_5_status"])
# ts_codes = ['002062.SZ', '002565.SZ']
for ts_code in ts_codes:
try:
start_date, end_date = date_now - datetime.timedelta(
days=90), date_now
security_point_data = SecurityData(
).get_qfq_security_point_data(ts_code, start_date, end_date)
buy_flag = calc_bs_data(security_point_data)
if buy_flag is True:
previous_work_day = Date().get_previous_workday(end_date)
previous_2_work_day = Date().get_previous_workday(
previous_work_day)
next_0_workday = date_now
next_1_workday = Date().get_next_workday(date_now)
next_2_workday = Date().get_next_workday(next_1_workday)
next_3_workday = Date().get_next_workday(next_2_workday)
next_4_workday = Date().get_next_workday(next_3_workday)
next_5_workday = Date().get_next_workday(next_4_workday)
security_point_data = SecurityData(
).get_qfq_security_point_data(
ts_code, previous_2_work_day,
end_date + datetime.timedelta(days=20))
security_daily_basic_data = SecurityData(
).get_security_daily_basic_data(
ts_code, previous_2_work_day,
end_date + datetime.timedelta(days=20))
security_data = pd.merge(security_point_data,
security_daily_basic_data,
on=["ts_code", "trade_date"])
security_data.drop(['close_y'], axis=1, inplace=True)
security_data.rename(index=str,
columns={"close_x": "close"},
inplace=True)
security_data.set_index(
security_daily_basic_data["trade_date"], inplace=True)
security_data = security_data.sort_index()
result_now = pd.DataFrame([
security_data.loc[previous_2_work_day],
security_data.loc[previous_work_day],
security_data.loc[next_0_workday],
security_data.loc[next_1_workday],
security_data.loc[next_2_workday],
security_data.loc[next_3_workday],
security_data.loc[next_4_workday],
security_data.loc[next_5_workday]
],
index=[[
'previous_2_day',
'previous_1_day',
'next_0_day', 'next_1_day',
'next_2_day', 'next_3_day',
'next_4_day', 'next_5_day'
], [1, 1, 1, 1, 1, 1, 1, 1]])
result_now = result_now.unstack(level=0)
result.append(result_now)
except Exception as e:
pass
return result
import datetime
from util.util_data.security_data import SecurityData
ts_code = "600030.SH"
security_point_data = SecurityData().get_security_point_data(ts_code, datetime.datetime(2016, 1, 1), datetime.datetime(2019, 3, 22))
security_daily_basic_data = SecurityData().get_security_daily_basic_data(ts_code, datetime.datetime(2016, 1, 1), datetime.datetime(2019, 3, 22))
security_point_data["turnover_rate"] = security_daily_basic_data["turnover_rate_f"]
security_point_data.drop("pre_close", axis=1, inplace=True)
security_point_data.drop("vol", axis=1, inplace=True)
security_point_data.drop("amount", axis=1, inplace=True)
security_point_data.drop("change", axis=1, inplace=True)
security_point_data.drop("ts_code", axis=1, inplace=True)
security_point_data.to_csv('./' + ts_code + '.csv', index=False, columns=['trade_date', 'open', 'low', 'high', 'close', 'turnover_rate', 'pct_chg'], header=False)
def idea_07(data):
"""
随机森林尝试
# https://www.cnblogs.com/zichun-zeng/p/4761602.html 模型保存
"""
max_depth_map_list = [
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
},
{
"next_2_day": 4,
"next_3_day": 1,
"next_4_day": 4,
"next_5_day": 1,
},
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
},
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 3,
}
]
data_low_low = data[
(data["open"]["next_1_day"] < data["close"]["next_0_day"]) & (data["close"]["next_1_day"] < data["close"]["next_0_day"]) & (
data["open"]["next_1_day"] >= data["close"]["next_1_day"])]
data_low_high = data[
(data["open"]["next_1_day"] < data["close"]["next_0_day"]) & (data["close"]["next_1_day"] > data["close"]["next_0_day"])]
data_high_low = data[
(data["open"]["next_1_day"] > data["close"]["next_0_day"]) & (data["close"]["next_1_day"] < data["close"]["next_0_day"])]
data_high_high = data[
(data["open"]["next_1_day"] > data["close"]["next_0_day"]) & (data["close"]["next_1_day"] > data["close"]["next_0_day"])]
for point_tendency_type_str, data, max_depth_map in zip(["data_low_low", "data_low_high", "data_high_low", "data_high_high"],
[data_low_low, data_low_high, data_high_low, data_high_high],
max_depth_map_list):
print(point_tendency_type_str)
percent_series = data["open"]["next_0_day"] / 100
target_day_list = ["next_2_day", "next_3_day", "next_4_day", "next_5_day"]
target_data_list = []
for day in target_day_list:
target_data_list.append([day, (data["close"][day] > data["high"]["next_1_day"]).values])
data = pd.DataFrame({
"open": data["open"]["next_0_day"] / percent_series,
"close": data["close"]["next_0_day"] / percent_series,
"high": data["high"]["next_0_day"] / percent_series,
"low": data["low"]["next_0_day"] / percent_series,
"turnover_rate_f": data["turnover_rate_f"]["next_0_day"],
"pct_chg": data["turnover_rate_f"]["next_0_day"],
}).values
forest_list = []
for day, target_data in target_data_list:
max_depth = max_depth_map[day]
X_train, X_test, y_train, y_test = train_test_split(data, target_data, random_state=0)
forest = RandomForestClassifier(n_estimators=100, random_state=0, max_depth=max_depth)
forest.fit(X_train, y_train)
forest_list.append([day, forest, X_train, X_test, y_train, y_test])
for day, forest, X_train, X_test, y_train, y_test in forest_list:
if forest.score(X_train, y_train) > 0.71 and forest.score(X_test, y_test) > 0.71:
print("day is {0}".format(day))
print("Accuracy on training set: {:.3f}".format(forest.score(X_train, y_train)))
print("Accuracy on test set: {:.3f}".format(forest.score(X_test, y_test)))
# all_ts_code = get_result_ts_code_list()
date_now = datetime.datetime.now()
date = datetime.datetime(date_now.year, date_now.month, date_now.day)
date = datetime.datetime(date_now.year, date_now.month, 24)
all_ts_code = Result().get_strategy_result_data("bbond", 'B', date)
for ts_code in all_ts_code:
security_point_data = SecurityData().get_security_point_data(ts_code, date, date)
security_daily_basic_data = SecurityData().get_security_daily_basic_data(ts_code, date, date)
security_data = pd.merge(security_point_data, security_daily_basic_data, on=["ts_code", "trade_date"])
security_data.drop(['close_y'], axis=1, inplace=True)
security_data.rename(index=str, columns={"close_x": "close"}, inplace=True)
security_data.set_index(security_daily_basic_data["trade_date"], inplace=True)
security_data = security_data.sort_index()
if security_data.empty is False:
security_data = security_data.loc[date]
else:
continue
predict_data = [security_data["open"], security_data["close"], security_data["high"],
security_data["low"]]
predict_data = np.array([predict_data]) / (predict_data[0] / 100)
predict_data = np.array([[predict_data[0][0], predict_data[0][1], predict_data[0][2], predict_data[0][3],
security_data["turnover_rate_f"], security_data["pct_chg"]]])
predict_result_now = forest.predict(predict_data)
if predict_result_now == [True]:
print(ts_code, predict_result_now)
# print(ts_code, predict_result_now)
print("\n\n")
print("\n\n\n")
def start():
security_data = SecurityData().get_index_point_data('399300.SZ', datetime.datetime(2016, 5, 3), datetime.datetime(2019, 1, 31))
# idea_07(security_data)
idea_08(security_data)
def idea_07(data):
"""
随机森林尝试
"""
max_depth_map_list = [
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
},
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
},
{
"next_2_day": 1,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
},
{
"next_2_day": 4,
"next_3_day": 1,
"next_4_day": 1,
"next_5_day": 1,
}
]
data_low_low = data[
(result["open"]["next_1_day"] < result["close"]["next_0_day"]) & (result["close"]["next_1_day"] < result["close"]["next_0_day"]) & (
result["open"]["next_1_day"] >= result["close"]["next_1_day"])]
data_low_high = data[
(result["open"]["next_1_day"] < result["close"]["next_0_day"]) & (result["close"]["next_1_day"] > result["close"]["next_0_day"])]
data_high_low = data[
(result["open"]["next_1_day"] > result["close"]["next_0_day"]) & (result["close"]["next_1_day"] < result["close"]["next_0_day"])]
data_high_high = data[
(result["open"]["next_1_day"] > result["close"]["next_0_day"]) & (result["close"]["next_1_day"] > result["close"]["next_0_day"])]
for point_tendency_type_str, data, max_depth_map in zip(["data_low_low", "data_low_high", "data_high_low", "data_high_high"],
[data_low_low, data_low_high, data_high_low, data_high_high],
max_depth_map_list):
print(point_tendency_type_str)
percent_series = data["open"]["next_0_day"] / 100
target_day_list = ["next_2_day", "next_3_day", "next_4_day", "next_5_day"]
target_data_list = []
for day in target_day_list:
target_data_list.append([day, (data["close"][day] > data["high"]["next_1_day"]).values])
data = pd.DataFrame({
"open": data["open"]["next_0_day"] / percent_series,
"close": data["close"]["next_0_day"] / percent_series,
"high": data["high"]["next_0_day"] / percent_series,
"low": data["low"]["next_0_day"] / percent_series
}).values
forest_list = []
for day, target_data in target_data_list:
max_depth = max_depth_map[day]
X_train, X_test, y_train, y_test = train_test_split(data, target_data, random_state=0)
forest = RandomForestClassifier(n_estimators=100, random_state=0, max_depth=max_depth)
forest.fit(X_train, y_train)
forest_list.append([day, forest, X_train, X_test, y_train, y_test])
for day, forest, X_train, X_test, y_train, y_test in forest_list:
if forest.score(X_train, y_train) > 0.71 and forest.score(X_test, y_test) > 0.71:
print("day is {0}".format(day))
print("Accuracy on training set: {:.3f}".format(forest.score(X_train, y_train)))
print("Accuracy on test set: {:.3f}".format(forest.score(X_test, y_test)))
# all_ts_code = get_result_ts_code_list()
date_now = datetime.datetime.now()
date = datetime.datetime(date_now.year, date_now.month, date_now.day)
all_ts_code = Result().get_strategy_result_data("a_p_5_p_10", 'B', date)
for ts_code in all_ts_code:
security_point_data = SecurityData().get_security_point_data(ts_code, date, date)
if security_point_data.empty is False:
security_point_data = security_point_data.loc[date]
else:
continue
predict_data = [security_point_data["open"], security_point_data["close"], security_point_data["high"],
security_point_data["low"]]
predict_data = np.array([predict_data]) / (predict_data[0] / 100)
predict_result_now = forest.predict(predict_data)
if predict_result_now == [True]:
print(ts_code, predict_result_now)
# print(ts_code, predict_result_now)
print("\n\n")
print("\n\n\n")