def filtration(self, model_for_filtration, choice_filter):
trend_filter = Trend() # Тренд фильтра
# Фильтр низких частот
if choice_filter == 1:
trend_filter.generation_trend_filter_potter()
# Фильтр высоких частот
if choice_filter == 2:
trend_filter.generating_trend_high_potter()
# Фильтр полосовой
if choice_filter == 3:
trend_filter.generating_trend_bandpass_filter()
# Фильтр режекторный
if choice_filter == 4:
trend_filter.generating_trend_notch_filter()
# Тренд существущей модели
trend_model = Trend()
trend_model.x = model_for_filtration.x
trend_model.y = model_for_filtration.y
trend_model.n = model_for_filtration.n
trend = convolution(trend_model, trend_filter)
self.x = trend.x
self.y = trend.y
def create_cut_table(self):
self.first_date = CandleTable.get_first_date(self.candle_table_name)
self.last_date = CandleTable.get_last_date(self.candle_table_name)
self.cut_trend_table_name = self.get_trend_table_name()
if DBManager.exists_table(self.cut_trend_table_name):
DBManager.drop_table(self.cut_trend_table_name)
tt = TrendTable(self.cut_trend_table_name)
tt.save()
candles = CandleTable.get_candle_array(self.candle_table_name)
for c in candles:
date = c.date
hits = TrendTable.get_most_recent_hits(self.trend_table_name, date)
trend = Trend(dbm, self.cut_trend_table_name, date, hits)
trend.save()
##cursor = TrendTable.get_section(self.trend_table_name, self.first_date, self.last_date)
##trend_tuples = cursor.fetchall()
##for t in trend_tuples:
##date = t[0]
##hits = t[1]
##trend = Trend(dbm, self.cut_trend_table_name, date, hits)
##trend.save()
##dbm.save_and_close()
return tt
def main():
db = MongoClient().test_database
n = 0.0
QUERY = read_in()
first = Trend()
first.setName(QUERY)
for tweet in db.tweets.find({
"entities.hashtags.text":
re.compile(QUERY, re.IGNORECASE)
}).sort([("_id", pymongo.DESCENDING)]):
trend_in_tweet = False
for hashtag in tweet["entities"]["hashtags"]:
if hashtag['text'].encode("utf-8").lower() != QUERY.lower():
first.addToRelated(hashtag['text'].encode("utf-8").lower(), 1)
n = n + 1
print QUERY.lower(), ": ", str(n)
top_10 = sorted(first.getRelated().items(),
key=operator.itemgetter(1),
reverse=True)
for top in top_10:
print top[0], ": ", float(top[1]) / n
first.clear()
def insert(self):
rows = self.data["table"]["rows"]
for r in rows:
try:
date_string = r["c"][0]["v"]
print(date_string)
date = date_to_timestamp(date_string)
hits = r["c"][1]["f"]
t = Trend(self.table_name, date, hits)
t.save()
except:
print("reached the end")
def get_trends(self) -> list:
trends = []
avarage_body_size = self.get_average_body_size_of_candles()
trend_candles = []
dir = self.candles[0].direction
for candle in self.candles:
if (candle.is_shot(avarage_body_size) or candle.direction == dir):
trend_candles.append(candle)
else:
trends.append(Trend(trend_candles))
trend_candles = [candle]
dir = candle.direction
return trends
def _new_trend(self, name):
if not self.__running: raise ENotRunning()
"""
Return an instance that implements ITrend interface for new trend with
no points.
"""
new_trend = Trend()
period = 60
points = []
preferences = {}
new_trend.configure({
'parent': self.trends,
'name': name,
'period': period,
'points': points,
'preferences': preferences
})
return new_trend
def convolution(trend_1, trend_2):
trend = Trend()
n = len(trend_1.y)
m = len(trend_2.y)
y = []
for i in range(n):
y_k = 0
for j in range(m):
coefficient_x = i - j
if coefficient_x >= 0:
y_m = trend_1.y[coefficient_x] * trend_2.y[j]
else:
# y_m = trend_2.y[j]
y_m = 0
y_k += y_m
y.append(y_k)
trend.y = np.array(y)
trend.x = trend_1.x
return trend
def runbacktest(begin, codename, dir, end, paint=True, star=False):
"""
设置策略环境参数，并运行部分回测策略
:param begin: 回测开始时间
:param codename: 回测股票代码或名称
:param dir: 作图目录地址
:param end: 回测结束时间，默认当前
:param paint: bool值，是否画图，默认为True
:param star: bool值，最后一个标记星星，默认为False
:return: 返回策略环境对象整体和图像地址
"""
print(f"BS日线策略：正在回测{codename}...")
try:
klist = ldd(begin, codename, end)
except loaddataError as e:
print(f"\t{e}")
raise loaddataError
print(f"\t{codename}获取日K线数{len(klist)}")
# 由于klist数据集是被hlp_env, trd_env和trade_env共享的，指向同一地址
# 因此任意指针均可调用
# 设置策略运行参数环境
hlp_env = HLPoint(klist, codename)
hlp_env.init_hl()
hlp_env.get_hl()
try:
trd_env = Trend(hlp_env)
trd_env.get_trend()
except TrendError as e:
print(e)
return
try:
trade_env = Trade(code=codename,
klist=hlp_env.klist,
hpi=hlp_env.hpi,
lpi=hlp_env.lpi)
trade_env.get_bs()
except Exception as e:
print(e)
return
for k in trade_env.klist:
k.t += " 15:00:00" # 统一日线时间与分钟级别时间的表示方法 "yyyy-mm-dd HH:MM:SS"
# note = [[1, hlp_env.temp_h, hlp_env.temp_min],
# [0, hlp_env.temp_l, hlp_env.temp_max]][hlp_env.hl == "l"]
locale.setlocale(locale.LC_CTYPE, "chinese")
# 创建回测表现统计操作对象
stat = Stat(hlp_env=hlp_env, trd_env=trd_env, trade_env=trade_env)
stat.get_mkt_p()
hlp_tb = stat.get_hl_info()
bsp_tb = stat.get_bs_info()
account_row = stat.get_account_info()
strategy_row = stat.get_strategy_info()
# 打包回测结果总结
# res_pac = {"hlp": hlp_tb, "bsp": bsp_tb, "account": account_row, "strategy": strategy_row}
stat.get_performance()
# 创建图像操作对象
graph_obj = BSgraph(trade_env=trade_env, stat=stat)
# 向回测全局对象返回结果总结
wrapper_df = graph_obj.wrapper_to_global_backtest()
# 该股票在总表中的信息
if paint:
# 绘图并获取图像地址
gdir = graph_obj.performance(
trdchg=trd_env.trdchg,
dir=dir,
star=star,
)
print(f"回测图像输出到{gdir}")
# return res_pac, gdir
return wrapper_df
else:
# return res_pac
return wrapper_df
def runbacktest(begin=BEGIN_DATE,
codename="600519.SH",
dir=GRAPH,
end=END_DATE,
paint=True,
star=False):
"""
设置策略环境参数，并运行部分回测策略
:param begin: 回测开始时间
:param codename: 回测股票代码或名称
:param thresh: 日线高低点间隔交易日长度
:param dir: 作图目录地址
:param end: 回测结束时间，默认当前
:param bounceThresh: 严格下跌过程中反弹超过bounceThresh交易日确认上涨，默认为65天
:param paint: bool值，是否画图，默认为True
:param star: bool值，最后一个标记星星，默认为False
:return: 返回策略环境对象整体和图像地址
"""
print(f"BS日线策略：正在回测{codename}...")
try:
klist = ldd(begin, codename, end)
except loaddataError as e:
print(f"\t{e}")
print(f"\t{codename}获取日K线数{len(klist)}")
# 设置策略运行参数环境
hlp_env = HLPoint(klist, codename)
hlp_env.init_hl()
hlp_env.get_hl()
try:
trd_env = Trend(hlp_env)
trd_env.get_trend()
except TrendError as e:
print(e)
return
note = [[1, hlp_env.temp_h, hlp_env.temp_min],
[0, hlp_env.temp_l, hlp_env.temp_max]][hlp_env.hl == "l"]
locale.setlocale(locale.LC_CTYPE, "chinese")
# 设置画图环境参数
graph_obj = BSgraph(hlist=hlp_env.hpi,
llist=hlp_env.lpi,
data=hlp_env.klist,
codename=codename,
note=note)
strategy_info = graph_obj.strategy_info(
max(hlp_env.hpi[-1], hlp_env.lpi[-1]), hlp_env.space_h,
hlp_env.space_l)
print(f"\t{codename}当前趋势{trd_env.trdnow}")
stock_info = []
for hh in hlp_env.hpi:
stock_info.append((hlp_env.klist[hh].t, "H", hlp_env.klist[hh].high,
hlp_env.klist[hlp_env.klist[hh].hl_confirmed].t))
for ll in hlp_env.lpi:
stock_info.append((hlp_env.klist[ll].t, "L", hlp_env.klist[ll].low,
hlp_env.klist[hlp_env.klist[ll].hl_confirmed].t))
stock_info.sort()
# 该股票在总表中的信息
if paint:
# 绘图并获取图像地址
gdir = graph_obj.performance(
trdchg=trd_env.trdchg,
dir=dir,
star=star,
)
return strategy_info, stock_info, gdir
else:
return strategy_info, stock_info
def filtration(self):
if self.radioButton_image.isChecked():
filter_trend = Trend()
type_of_filter = self.comboBox_3.currentText()
m = self.lineEdit_2.text()
delta_t = self.lineEdit.text()
fc_1 = self.lineEdit_3.text()
fc_2 = self.lineEdit_4.text()
error_dialog = QtWidgets.QErrorMessage()
if m != '':
try:
m = int(m)
filter_trend.m = m
except:
error_dialog.showMessage(
'Значение m должны быть целочисленными. Использовано значение m по '
'умолчанию')
if delta_t != '':
try:
delta_t = float(delta_t)
filter_trend.delta_t = delta_t
except:
error_dialog.showMessage(
'Значение delta t должны быть вещественными. Использовано значение delta t по '
'умолчанию')
if fc_1 != '':
try:
fc_1 = int(fc_1)
filter_trend.fc_1 = fc_1
except:
error_dialog.showMessage(
'Значение fc_1 должны быть целочисленными. Использовано значение fc_1 по '
'умолчанию')
if fc_2 != '':
try:
fc_2 = int(fc_2)
filter_trend.fc_2 = fc_2
except:
error_dialog.showMessage(
'Значение fc_2 должны быть целочисленными. Использовано значение fc_2 по '
'умолчанию')
if type_of_filter == 'Низких частот':
filter_trend.generation_trend_filter_potter()
elif type_of_filter == 'Высоких частот':
filter_trend.generating_trend_high_potter()
elif type_of_filter == 'Полосовой':
filter_trend.generating_trend_bandpass_filter()
elif type_of_filter == 'Режекторный':
filter_trend.generating_trend_notch_filter()
place_to_show_image: int = int(self.comboBox_2.currentText())
try:
position_img: int = int(self.comboBox.currentText())
img_path: str = POSITION_FOR_ANALYSIS.get(position_img)
except:
error_dialog.showMessage(
'Не найдено изображение для фильтрации')
return
filtration(img_path, filter_trend, place_to_show_image)
else:
type_of_filter_with_mask: str = self.comboBox_4.currentText()
place_to_show_image: int = int(self.comboBox_2.currentText())
error_dialog: object = QtWidgets.QErrorMessage()
try:
n_of_maska: int = int(self.lineEdit_5.text())
if n_of_maska % 2 == 0:
n_of_maska = n_of_maska + 1
except:
n_of_maska: int = 3
try:
position_img: int = int(self.comboBox.currentText())
img_path: str = POSITION_FOR_ANALYSIS.get(position_img)
except:
error_dialog.showMessage(
'Не найдено изображение для фильтрации')
return
if type_of_filter_with_mask == 'Арифметический':
arithmetic_filter_img(img_path, place_to_show_image,
n_of_maska)
elif type_of_filter_with_mask == 'Медианный':
median_img(img_path, place_to_show_image, n_of_maska)
self.main_window.show_img(place_to_show_image)
self.close_window()
import operator
import math
import getopt
def take(n, iterable):
"Return first n items of the iterable as a list"
return list(islice(iterable, n))
db = MongoClient().test_database
n = 0.0
QUERY = 'ROOT'
first = Trend()
first.setName(QUERY)
for tweet in db.tweets.find():
for hashtag in tweet["entities"]["hashtags"]:
if hashtag['text'].encode("utf-8").lower() != QUERY.lower():
first.addToRelated(hashtag['text'].encode("utf-8").lower(), 1)
n = n + 1
print QUERY.lower(), ": ", str(n)
top_10 = sorted(first.getRelated().items(),
key=operator.itemgetter(1),
reverse=True)
top_10 = take(20, top_10)
for top in top_10:
print "\t", top[0], ": ", float(top[1]) / n
def get_trends_in_place(now, woeid, history=False):
'''
Get trends in place, current date and place (woeid)
https://developer.twitter.com/en/docs/trends/trends-for-location/api-reference/get-trends-place
'''
queryResult = Location.query(Location.woeid == int(woeid)).fetch(1)
location = (queryResult[0] if len(queryResult) == 1 else None)
# check for history
trend_history = None
if location is not None:
queryResult = TrendHistory.query(
ndb.AND(TrendHistory.date == now,
TrendHistory.location == location)).fetch(1)
trend_history = (queryResult[0] if len(queryResult) == 1 else None)
if trend_history is not None:
result = []
result.append(trend_history.to_dict())
return result
# Only search at history
if (history):
return ""
# If data is not in cache do request
if not twitter_api.isAuthenticated():
twitter_api.auth()
result = json.loads(twitter_api.get_trends_place(woeid))
if (isinstance(result, list)):
# persist result
trends = []
for twitter_trend in result[0]['trends']:
queryResult = Trend.query(
Trend.name == twitter_trend['name']).fetch(1)
trend = (queryResult[0] if len(queryResult) == 1 else Trend())
# FIXME(ruben): check keys before assignment
trend.name = twitter_trend['name']
trend.promoted_content = twitter_trend['promoted_content']
trend.query_string = twitter_trend['query']
trend.url = twitter_trend['url']
trend.tweet_volume = twitter_trend['tweet_volume']
# increases search count, if trend already exists
trend.search_count = (trend.search_count +
1 if len(queryResult) == 1 else 1)
trends.append(trend)
if len(trends) > 0:
ndb.put_multi(trends)
if location is None:
location = Location()
location.name = result[0]['locations'][0]['name']
location.woeid = result[0]['locations'][0]['woeid']
location.put()
# create trend history
trend_history = TrendHistory()
trend_history.date = now
trend_history.trends = trends
trend_history.location = location
trend_history.put()
return result
else:
print("Error:", result)
return ""
def sum_trend(trend_1, trend_2):
trend = Trend()
trend.y = trend_1.y + trend_2.y
trend.x = trend_1.x
return trend
def multi(trend_1, trend_2):
trend = Trend()
trend.y = trend_1.y * trend_2.y
trend.x = trend_1.x
return trend
def calculation(self):
# y(x)=kx+b
if self.option == 1:
trend = Trend()
trend.generating_trend_line()
self.y = trend.y
# y(x)=-kx+b
if self.option == 2:
trend = Trend()
trend.generating_trend_line()
trend.y = np.flip(trend.y)
self.y = trend.y
# y(x) = beta * exp^(alpha * i)
if self.option == 3:
trend = Trend()
trend.generating_exhibitor()
self.y = trend.y
# y(x) = beta * exp^(alpha * -i)
if self.option == 4:
trend = Trend()
trend.generating_exhibitor()
trend.y = np.flip(trend.y)
self.y = trend.y
# Встроенный рандом
if self.option == 5:
trend = Trend()
trend.generating_trend_random(self.s_min, self.s_max)
self.y = trend.y
# Кастомный рандом
if self.option == 6:
trend = Trend()
trend.generating_custom_random()
self.y = trend.y
# Рандом + сдвиг
if self.option == 7:
trend1 = Trend()
trend1.generating_trend_random(self.s_min, self.s_max)
trend = self.shift(trend1)
self.y = trend.y
# Указали, что не требуется нормализация
self.flag_normalisation = 0
# Значения за областью
if self.option == 8:
trend1 = Trend()
trend1.generating_random_spikes(self.s_min, self.s_max)
self.y = trend1.y
# Указали, что не требуется нормализация
self.flag_normalisation = 0
# Адитивная модель №1
if self.option == 9:
trend1 = Trend()
trend1.generating_trend_line()
trend1.y = np.flip(trend1.y)
trend2 = Trend()
trend2.generating_trend_random(self.s_min, self.s_max)
trend = sum_trend(trend1, trend2)
self.y = trend.y
# Адитивная модель №2
if self.option == 10:
trend1 = Trend()
trend1.generating_trend_line()
trend2 = Trend()
trend2.generating_trend_random(self.s_min, self.s_max)
trend = sum_trend(trend1, trend2)
self.y = trend.y
# Мультипликативная модель №1
if self.option == 11:
trend1 = Trend()
trend1.generating_trend_line()
trend1.y = np.flip(trend1.y)
trend2 = Trend()
trend2.generating_trend_random(self.s_min, self.s_max)
trend = multi(trend1, trend2)
self.y = trend.y
# Мультипликативная модель №2
if self.option == 12:
trend1 = Trend()
trend1.generating_trend_line()
trend2 = Trend()
trend2.generating_trend_random(self.s_min, self.s_max)
trend = multi(trend1, trend2)
self.y = trend.y
# График кусочной функции
if self.option == 13:
trend = Trend()
trend.generating_piecewise_function(self.s_min, self.s_max)
self.y = trend.y
# График гармонический процесс
if self.option == 17:
trend = Trend()
trend.generating_harmonic_process()
if self.c != 0:
self.y = self.y + self.c
else:
self.y = trend.y
self.flag_normalisation = 0
self.normalisation_axis()
# График полигармонического процесса
# x(t) = x1(t) + x2(t) = x3(t)
# xi(t) = Ai * sin(2piFit)
# A1 = 25 f1 = 11
# A2 = 35 f2 = 41
# A3 = 30 f3 = 141
if self.option == 19:
trend1 = Trend()
trend2 = Trend()
trend3 = Trend()
trend1.a_0 = 25
trend1.f_0 = 11
trend1.generating_harmonic_process()
trend2.a_0 = 35
trend2.f_0 = 41
trend2.generating_harmonic_process()
trend3.a_0 = 30
trend3.f_0 = 141
trend3.generating_harmonic_process()
trend4 = sum_trend(trend1, trend2)
trend = sum_trend(trend4, trend3)
self.y = trend.y
# График Рандом + спайки
if self.option == 20:
trend_1 = Trend()
trend_1.generating_trend_random(self.s_min, self.s_max)
trend_2 = Trend()
trend_2.generating_random_spikes(self.s_min, self.s_max)
trend = sum_trend(trend_1, trend_2)
self.y = trend.y
self.flag_normalisation = 0
self.axis_max = np.amax(self.y) * 1.2
self.axis_min = np.amin(self.y) * 1.2
# График Гармонический процесс + trend
if self.option == 25:
trend_1 = Trend()
trend_2 = Trend()
trend_1.generating_harmonic_process()
trend_2.generating_trend_line()
trend = sum_trend(trend_1, trend_2)
self.y = trend.y
# График Гармонический процесс + спайки
if self.option == 26:
trend_1 = Trend()
trend_2 = Trend()
trend_1.generating_harmonic_process()
trend_2.generating_random_spikes(self.s_min, self.s_max)
trend = sum_trend(trend_1, trend_2)
self.y = trend.y
self.flag_normalisation = 0
self.normalisation_axis()
# График ГП(гармонический процесс) + спайки + рандом + trend
if self.option == 27:
trend_1 = Trend()
trend_2 = Trend()
trend_3 = Trend()
trend_4 = Trend()
trend_1.generating_harmonic_process()
trend_2.generating_random_spikes(self.s_min, self.s_max)
trend_3.generating_trend_random(self.s_min, self.s_max)
trend_4.generating_trend_line()
trend = (trend_1, trend_2)
trend = (trend, trend_3)
trend = (trend, trend_4)
self.y = trend.y
self.flag_normalisation = 0
self.normalisation_axis()
# График из файла
if self.option == 28:
trend = Trend()
trend.generating_trend_from_file()
self.y = trend.y
self.flag_normalisation = 0
self.normalisation_axis()
# График ГП + экспонента (кардиограма)
if self.option == 29:
self.n = 200
self.x = np.arange(0, self.n)
# self.delta_t = 0.005
# self.display_n = self.n
self.s_max = 1
trend_1 = Trend()
trend_1.n = self.n
trend_1.x = self.x
trend_1.delta_t = 0.005
trend_1.generating_harmonic_process()
trend_2 = Trend()
trend_2.n = self.n
trend_2.x = self.x
trend_2.generating_exhibitor()
# Получили тренд сердцебиения
trend = multi(trend_1, trend_2)
# self.y = trend.y
# self.normalization()
# trend.y = self.y
self.n = 1000
trend_3 = Trend()
trend_3.generating_spikes(100)
trend = convolution(trend_3, trend)
self.y = trend.y
self.x = trend.x
self.s_max = 100
self.s_min = -self.s_max
# Реализация фильтров
# Низких частот
if self.option == 30:
trend = Trend()
trend.generation_trend_filter_potter()
self.y = trend.y
self.x = trend.x
self.n = trend.n
self.display_n = trend.display_n
self.dt = trend.dt
# Фильтр высоких частот
if self.option == 31:
trend = Trend()
trend.generating_trend_high_potter()
self.y = trend.y
self.x = trend.x
self.n = trend.n
self.display_n = trend.display_n
self.dt = trend.dt
# Полосовой фильтр
if self.option == 32:
trend = Trend()
trend.generating_trend_bandpass_filter()
self.y = trend.y
self.x = trend.x
self.n = trend.n
self.display_n = trend.display_n
self.dt = trend.dt
# Режекторный фильтр
if self.option == 33:
trend = Trend()
trend.generating_trend_notch_filter()
self.y = trend.y
self.x = trend.x
self.n = trend.n
self.display_n = trend.display_n
self.dt = trend.dt
# Звук ma.wav
if self.option == 34:
self.name_of_wav_file = "../input files/custom.wav"
sound_trend = Sound(self.name_of_wav_file)
self.x = sound_trend.x
self.y = sound_trend.y * self.c
self.rate = sound_trend.rate
self.n = len(self.x)
self.flag_checking_display_n = 1
# Звук my_voice.wav
if self.option == 35:
self.name_of_wav_file = "../input files/my_voice.wav"
sound_trend = Sound(self.name_of_wav_file)
self.x = sound_trend.x
self.y = sound_trend.y * self.c
self.rate = sound_trend.rate
self.n = len(self.x)
self.flag_checking_display_n = 1
# Экзамен
if self.option == 36:
trend = Trend()
trend.generating_trend_from_file_exam()
self.y = trend.y
self.flag_normalisation = 0
self.normalisation_axis()
# Модель Input: кардиограма, заполненная нулями с 200 до 1000
if self.option == 37:
self.n = 200
self.x = np.arange(0, self.n)
self.s_max = 1
trend_1 = Trend()
trend_1.n = self.n
trend_1.x = self.x
trend_1.delta_t = 0.005
trend_1.generating_harmonic_process()
trend_2_n = 800
trend_2_y = np.zeros(trend_2_n)
self.n = 1000
self.x = np.arange(0, self.n)
self.y = np.concatenate([trend_1.y, trend_2_y])
print(self.y)
self.s_max = 100
self.s_min = -self.s_max
