def addTechnicalIndicators(simple_data):
stock = sts.StockDataFrame(simple_data)
stock['cr']
stock['kdjk']
stock['kdjd']
stock['kdjj']
stock['close_10_sma']
stock['macd']
stock['boll']
stock['rsi_6']
stock['rsi_12']
stock['wr_6']
stock['wr_10']
stock['cci']
stock['cci_20']
stock['tr']
stock['atr']
stock['dma']
stock['adxr']
stock['close_12_ema']
stock['trix']
stock['trix_9_sma']
stock['vr']
stock['vr_6_sma']
new_dataframe = pd.DataFrame(stock).loc[:, COLUMNS]
transformed_dataframe = new_dataframe.iloc[
10:len(new_dataframe) - SHIFT_NUMBER] # Bỏ các hàng khuyết dữ liệu
scaler = StandardScaler()
scaler.fit(transformed_dataframe)
return pd.DataFrame(scaler.transform(transformed_dataframe),
columns=COLUMNS)
def calculate_stock(self):
stock = stockstats.StockDataFrame().retype(self.feed)
stock['close_20_sma']
stock['boll']
stock['boll_ub']
stock['boll_lb']
return stock.round(2)
def calculate_stock(self):
stock = stockstats.StockDataFrame().retype(self.feed)
shortlen = 5
longlen = 10
short = talib.EMA(self.feed['volume'], shortlen)
long = talib.EMA(self.feed['volume'], longlen)
osc = 100 * (short - long) / long
stock['VOSC'] = osc
return stock
def __init__(self, df: pd.core.frame.DataFrame):
# Arg Check
co = ['open', 'high', 'low', 'close']
if not all(i in df.columns for i in co):
raise KeyError("arg's columns must have {}, but it has {}".format(co, df.columns))
self._init_stock_dataframe = ss.StockDataFrame(df) # スパン変更前のデータフレーム
self.stock_dataframe = None # スパン変更後、インジケータ追加後のデータフレーム
self.freq = None # 足の時間幅
self._fig = None # <-- plotly.graph_objs
self._indicators = {} # Plotするときに使う指標
def __init__(self, df: pd.core.frame.DataFrame):
# Arg Check
co = ['open', 'high', 'low', 'close']
assert all(
i in df.columns
for i in co), 'arg\'s columns must have {}, but it has {}'.format(
co, df.columns)
if not type(df.index) == pd.tseries.index.DatetimeIndex:
raise TypeError(df.index)
self._init_stock_dataframe = ss.StockDataFrame(df) # スパン変更前のデータフレーム
self.stock_dataframe = None # スパン変更後、インジケータ追加後のデータフレーム
self.freq = None # 足の時間幅
self._fig = None # <-- plotly.graph_objs
def resample(self, freq: str):
"""Convert ohlc time span
Usage: `fx.resample('D') # 日足に変換`
* Args: 変更したい期間 M(onth) | W(eek) | D(ay) | H(our) | T(Minute) | S(econd)
* Return: スパン変更後のデータフレーム
"""
self.freq = freq
df = self._init_stock_dataframe.resample(freq).ohlc2().dropna()
self.stock_dataframe = ss.StockDataFrame(df)
for indicator in self._indicators.keys():
self.append(indicator) # Re-append indicator in dataframe
return self.stock_dataframe
def calulate_stock(self):
#change class attributes before calculation
stockstats.StockDataFrame.BOLL_PERIOD = self.boll_period
stock = stockstats.StockDataFrame().retype(self.feed)
stock['boll']
#stock._get_boll(stockstats.StockDataFrame())
stock['boll_ub']
stock['boll_lb']
#print('period',stockstats.StockDataFrame.BOLL_PERIOD)
#print(stock['boll_lb'].tail(3))
stock['close_30_sma']
#boll limit
stock['boll_bb'] = (stock['close'] - stock['boll_lb'])/(stock['boll_ub']-stock['boll_lb'])*100
stock['boll_band'] = (stock['boll_ub'] - stock['boll_lb'])/stock['boll']*100
#print(stock.tail(5))
#print(stock['boll'].tail(3))
#print(ind.up_trend(stock['boll'],window=2))
#print(ind.down_trend(stock['boll'],window=1))
return stock.round(2)
def run(self):
print(f"Process_{self.no}_started")
self.stats = stockstats.StockDataFrame(self.data)
def normalize(x):
x = x.copy()
return (x - np.mean(x)) / (np.std(x) + 1e-5)
self.wr = normalize(self.stats['wr_14'])
self.kdjk = normalize(self.stats['kdjk_20'])
self.kdjd = normalize(self.stats['kdjd_20'])
self.kdjj = normalize(self.stats['kdjj_20'])
self.rsi = normalize(self.stats['rsi_18'])
self.adx = normalize(self.stats['adx'])
self.dma = normalize(self.stats['dma'])
self.len = len(self.stats)
def shirnk(x):
if abs(x) > self.threshold:
return x
else:
return 0
sign2cate = lambda x: 1 - x
while True:
pos = np.random.randint(self.img_shape[1], self.len - 1 - self.k)
img = self.grey_scale_img(pos)
delta = self.stats.iloc[
pos + self.k]['close'] - self.stats.iloc[pos]['close']
t = shirnk(delta)
sign = np.sign(t)
cate = sign2cate(sign)
stat = self.get_stats(pos)
self.pair_queue.put((img, stat, cate))
def __init__(self, data, total_step, img_shape, charge, init_money,
threshold):
self.stats = stockstats.StockDataFrame(data)
def normalize(x):
x = x.copy()
return (x - np.mean(x)) / (np.std(x) + 1e-5)
self.wr = normalize(self.stats['wr_14'])
self.kdjk = normalize(self.stats['kdjk_20'])
self.kdjd = normalize(self.stats['kdjd_20'])
self.kdjj = normalize(self.stats['kdjj_20'])
self.rsi = normalize(self.stats['rsi_12'])
self.adx = normalize(self.stats['adx'])
self.dma = normalize(self.stats['dma'])
self.len = len(self.stats)
self.start_pos = None
self.cur_pos = None
self.total_step = total_step
self.step_cnt = None
self.in_cnt = None
self.is_win = None
self.reward_list = None
self.img_shape = img_shape
self.charge = charge
self.action = None # 0:buy, 1:wait, 2:sell
self.f = None
self.threshold = threshold
self.init_money = init_money
self.money = None
self.contract = None
self.sign = None
self.in_close = None
self.window = None
print("Env_started")
def calculate_long(self):
data = ind.load_stock(self.tker,200)
return stockstats.StockDataFrame().retype(data)
def get_technical_indicators(data, last_days=2691, plot=False):
# 获取股价技术指标 输入参数为数据集、持续时间和是否绘制图表 输出技术指标 key表示对哪一个指标进行统计分析
# 7日均线和21日均线
import stockstats
dataset_tech = data[[
'open', 'close_x_x', 'high', 'low', 'vol', 'amount', 'pct_chg'
]]
dataset_tech['close'] = dataset_tech['close_x_x']
dataset_tech['volume'] = dataset_tech['vol']
dataset_tech = dataset_tech.drop(columns=['close_x_x', 'vol'])
stock = stockstats.StockDataFrame(dataset_tech)
technical_keys = [
'macd', # moving average convergence divergence. Including signal and histogram.
'macds', # MACD signal line
'macdh', # MACD histogram
'volume_delta', # volume delta against previous day
'open_2_d', # open delta against next 2 day
'open_-2_r', # open price change (in percent) between today and the day before yesterday, 'r' stands for rate.
'cr', # CR indicator, including 5, 10, 20 days moving average
'cr-ma1',
'cr-ma2',
'cr-ma3',
'volume_-3,2,-1_max', # volume max of three days ago, yesterday and two days later
'volume_-3~1_min', # volume min between 3 days ago and tomorrow
'kdjk', # KDJ, default to 9 days
'kdjd',
'kdjj',
'kdjk_3_xu_kdjd_3', # three days KDJK cross up 3 days KDJD
'open_2_sma', # 2 days simple moving average on open price
'boll', # bolling, including upper band and lower band
'boll_ub',
'boll_lb',
'close_10.0_le_5_c', # close price less than 10.0 in 5 days count
'cr-ma2_xu_cr-ma1_20_c', # CR MA2 cross up CR MA1 in 20 days count
'rsi_6', # 6 days RSI
'rsi_12', # 12 days RSI
'wr_10', # 10 days WR
'wr_6', # 6 days WR
'cci', # CCI, default to 14 days
'cci_20', ## 20 days CCI
'tr', #TR (true range)
'atr', # ATR (Average True Range)
'dma', # DMA, difference of 10 and 50 moving average
'pdi', # DMI +DI, default to 14 days
'mdi', # -DI, default to 14 days
'dx', # DX, default to 14 days of +DI and -DI
'adx', # ADX, 6 days SMA of DX, same as stock['dx_6_ema']
'adxr', # ADXR, 6 days SMA of ADX, same as stock['adx_6_ema']
'trix', # TRIX, default to 12 days
'trix_9_sma', # MATRIX is the simple moving average of TRIX
'vr', # VR, default to 26 days
'vr_6_sma' # MAVR is the simple moving average of VR
]
for key in technical_keys:
dataset_tech[key] = pd.DataFrame(stock[key])
if plot: # 绘制技术指标
plot_dataset = dataset_tech
plot_dataset = dataset_tech.iloc[-last_days:, :]
shape_0 = plot_dataset.shape[0]
x = list(plot_dataset.index)
plt.figure(dpi=100)
# 1.股价、成交量，成交额、MA移动平均线、MACD
plt.subplot(2, 1, 1)
plt.plot(plot_dataset['close'], label='Close Price')
plt.xticks(plot_dataset['volume'])
plt.plot(plot_dataset[''])
# 2.KDJ
# 3.RSI
# 4.BOLL
# 5.WR
# 6.DMI
# 7.All in one
for key in technical_keys:
plt.plot(dataset_tech[key], label=key, color=choose_color())
plt.legend()
plt.show()
# Plot first subplot
plt.subplot(2, 1, 1)
plt.plot(plot_dataset[key + '_' + 'tech'],
label=key + '_' + 'technical indicator',
linestyle='-')
plt.plot(plot_dataset[key + '_' + 'ma7'],
label=key + '_' + 'MA 7',
color=colors[0],
linestyle='--')
plt.plot(plot_dataset[key + '_' + 'ma21'],
label=key + '_' + 'MA 21',
color=colors[1],
linestyle='--')
plt.plot(plot_dataset[key + '_' + '20sd'],
label=key + '_' + '20days Standard Deviation',
color=colors[6],
linestyle='--')
plt.plot(plot_dataset[key + '_' + 'upper_band'],
label=key + '_' + 'Upper Band',
color=colors[2],
linestyle=':')
plt.plot(plot_dataset[key + '_' + 'lower_band'],
label=key + '_' + 'Lower Band',
color=colors[3],
linestyle=':')
plt.fill_between(x,
plot_dataset[key + '_' + 'lower_band'],
plot_dataset[key + '_' + 'upper_band'],
alpha=0.3)
plt.title(' Technical indicators in {} days .'.format(last_days))
plt.ylabel(key + '_' + 'technical')
plt.legend()
# Plot second subplot
plt.subplot(2, 1, 2)
plt.title('MACD')
plt.plot(plot_dataset[key + '_' + 'MACD'],
label=key + '_' + 'MACD',
linestyle='-')
plt.hlines(plot_dataset[key + '_' + 'MACD'].mean(),
plot_dataset.index[0],
plot_dataset.index[shape_0 - 1],
color=colors[4],
linestyles=':',
label='MACD mean')
plt.plot(plot_dataset[key + '_' + 'ema'],
label=key + '_' + 'Ema',
linestyle='--')
plt.hlines(plot_dataset[key + '_' + 'ema'].mean(),
plot_dataset.index[0],
plot_dataset.index[shape_0 - 1],
colors=colors[5],
linestyles=':',
label='Exponential moving average mean')
plt.legend()
plt.show()
# 汇总技术指标
return dataset_tech
def reset_dataframe(df):
"""Reset dataframe as stockstats"""
return ss.StockDataFrame(df.ix[:, ['open', 'high', 'low', 'close']])
def calulate_stock(self):
stock = stockstats.StockDataFrame().retype(self.feed)
#print(stock['boll'].tail(3))
return stock.round(2)
def get_technical_indicators(data, last_days=2691, plot=True, save=False):
# 获取股价技术指标 输入参数为数据集、持续时间和是否绘制图表 输出技术指标 key表示对哪一个指标进行统计分析
# 7日均线和21日均线
import stockstats
dataset_tech = data[[
'daily_open', 'daily_close', 'daily_high', 'daily_low', 'daily_vol',
'daily_amount'
]]
dataset_tech = dataset_tech.rename(
columns=lambda x: x.lstrip('daily_')).rename(columns={
'vol': 'volume',
'ow': 'low',
'mount': 'amount'
})
stock = stockstats.StockDataFrame(dataset_tech)
technical_keys = [
'macd', # moving average convergence divergence. Including signal and histogram.
'macds', # MACD signal line
'macdh', # MACD histogram
'volume_delta', # volume delta against previous day
'volume_-3,2,-1_max', # volume max of three days ago, yesterday and two days later
'volume_-3~1_min', # volume min between 3 days ago and tomorrow
'kdjk', # KDJ, default to 9 days
'kdjd',
'kdjj',
'kdjk_3_xu_kdjd_3', # three days KDJK cross up 3 days KDJD
'boll', # bolling, including upper band and lower band
'boll_ub',
'boll_lb',
'open_2_sma', # 2 days simple moving average on open price
'open_2_d', # open delta against next 2 day
'open_-2_r', # open price change (in percent) between today and the day before yesterday, 'r' stands for rate.
'close_10.0_le_5_c', # close price less than 10.0 in 5 days count
'cr', # CR indicator, including 5, 10, 20 days moving average
'cr-ma1',
'cr-ma2',
'cr-ma3',
'cr-ma2_xu_cr-ma1_20_c', # CR MA2 cross up CR MA1 in 20 days count
'rsi_6', # 6 days RSI
'rsi_12', # 12 days RSI
'wr_10', # 10 days WR
'wr_6', # 6 days WR
'cci', # CCI, default to 14 days
'cci_20', ## 20 days CCI
'dma', # DMA, difference of 10 and 50 moving average
'pdi', # DMI +DI, default to 14 days
'mdi', # -DI, default to 14 days
'dx', # DX, default to 14 days of +DI and -DI
'adx', # ADX, 6 days SMA of DX, same as stock['dx_6_ema']
'adxr', # ADXR, 6 days SMA of ADX, same as stock['adx_6_ema']
'tr', #TR (true range)
'atr', # ATR (Average True Range)
'trix', # TRIX, default to 12 days
'trix_9_sma', # MATRIX is the simple moving average of TRIX
'vr', # VR, default to 26 days
'vr_6_sma' # MAVR is the simple moving average of VR
]
for key in technical_keys:
dataset_tech[key] = pd.DataFrame(stock[key])
dataset_tech['ma7'] = dataset_tech['close'].rolling(window=7).mean()
dataset_tech['ma21'] = dataset_tech['close'].rolling(window=21).mean()
dataset_tech['ema'] = dataset_tech['close'].ewm(com=0.5).mean()
dataset_tech['momentum'] = dataset_tech['close'] - 1
if plot: # 绘制技术指标
plot_dataset = dataset_tech
plot_dataset = dataset_tech.iloc[-last_days:, :]
shape_0 = plot_dataset.shape[0]
x = list(plot_dataset.index)
colors = choose_color(10)
# 0.股价、成交量，成交额、MA移动平均线
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(3, 1, 1)
plt.title('Close Price and Volume Statistics')
plt.plot(plot_dataset['close'], label='Close Price')
plt.plot(plot_dataset['ma7'],
label='MA-7',
linestyle='--',
linewidth=linewidth)
plt.plot(plot_dataset['ma21'],
label='MA-21',
linestyle='--',
linewidth=linewidth)
plt.plot(plot_dataset['ema'],
label='EMA',
linestyle=':',
linewidth=linewidth)
plt.legend()
plt.subplot(3, 1, 2)
plt.bar(x, plot_dataset['volume'], label='Volume', width=linewidth)
plt.bar(x, -plot_dataset['amount'], label='Amount', width=linewidth)
plt.legend()
plt.subplot(3, 1, 3)
plt.plot(plot_dataset['volume_delta'],
label='volume delta',
linestyle='-',
linewidth=linewidth / 2,
color='k')
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\30_price_amount.png')
plt.show()
# 1.MACD
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(2, 1, 1)
plt.title('Close price and MACD')
plt.plot(plot_dataset['close'], label='Close Price')
plt.legend()
plt.subplot(2, 1, 2)
plt.plot(plot_dataset['macd'], label='macd', linewidth=linewidth)
plt.plot(plot_dataset['macds'],
label='macd signal line',
linestyle='--',
linewidth=linewidth)
plt.bar(plot_dataset['macdh'].loc[plot_dataset['macdh'] >= 0].index,
plot_dataset['macdh'].loc[plot_dataset['macdh'] >= 0],
label='macd histgram',
width=linewidth,
color='r')
plt.bar(plot_dataset['macdh'].loc[plot_dataset['macdh'] < 0].index,
plot_dataset['macdh'].loc[plot_dataset['macdh'] < 0],
label='macd histgram',
width=linewidth,
color='g')
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\31_MACD.png')
plt.show()
# 2.KDJ and BOLL
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(2, 1, 1)
plt.title('Bolling band and KDJ')
plt.plot(plot_dataset['close'], label='Close Price')
plt.plot(plot_dataset['boll'],
label='Bolling',
linestyle='--',
linewidth=linewidth)
plt.plot(plot_dataset['boll_ub'],
color='c',
label='Bolling up band',
linewidth=linewidth)
plt.plot(plot_dataset['boll_lb'],
color='c',
label='Bolling low band',
linewidth=linewidth)
plt.fill_between(x,
plot_dataset['boll_ub'],
plot_dataset['boll_lb'],
alpha=0.35)
plt.legend()
plt.subplot(2, 1, 2)
plt.plot(plot_dataset['kdjk'], label='KDJ-K', linewidth=linewidth)
plt.plot(plot_dataset['kdjd'], label='KDJ-K', linewidth=linewidth)
plt.plot(plot_dataset['kdjj'], label='KDJ-K', linewidth=linewidth)
plt.scatter(
plot_dataset['kdjk'].loc[plot_dataset['kdjk_3_xu_kdjd_3'] ==
True].index,
plot_dataset['kdjk'].loc[plot_dataset['kdjk_3_xu_kdjd_3'] == True],
marker='^',
color='r',
label='three days KDJK cross up 3 days KDJD')
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\32_boll_kdj.png')
plt.show()
# 3.Open price and RSI
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(2, 1, 1)
plt.title('Open price and RSI')
plt.plot(plot_dataset['open'], label='Open Price')
plt.bar(x,
plot_dataset['open_2_d'],
label='open delta against next 2 day')
plt.plot(
plot_dataset['open_-2_r'],
label=
'open price change (in percent) between today and the day before yesterday',
linewidth=linewidth)
plt.legend()
plt.subplot(2, 1, 2)
plt.plot(plot_dataset['rsi_12'], label='12 days RSI ', color='c')
plt.plot(plot_dataset['rsi_6'],
label='6 days RSI',
linewidth=linewidth,
color='r')
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\33_open_rsi.png')
plt.show()
# 4.CR and WR
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(2, 1, 1)
plt.title('WR and CR in 5/10/20 days')
plt.plot(plot_dataset['wr_10'],
label='10 days WR',
linestyle='-',
linewidth=linewidth,
color='g')
plt.plot(plot_dataset['wr_6'],
label='6 days WR',
linestyle='-',
linewidth=linewidth,
color='r')
plt.legend()
plt.subplot(2, 1, 2)
plt.bar(x,
plot_dataset['cr'],
label='CR indicator',
linestyle='--',
linewidth=linewidth,
color='skyblue')
plt.plot(plot_dataset['cr-ma1'],
label='CR 5 days MA',
linestyle='-',
linewidth=linewidth)
plt.plot(plot_dataset['cr-ma2'],
label='CR 10 days MA',
linestyle='-',
linewidth=linewidth)
plt.plot(plot_dataset['cr-ma3'],
label='CR 20 days MA',
linestyle='-',
linewidth=linewidth)
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\34_cr_ma.png')
plt.show()
# 5.CCI TR VR
#
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(2, 1, 1)
plt.title('CCI TR and VR')
plt.plot(plot_dataset['tr'],
label='TR (true range)',
linewidth=linewidth)
plt.plot(plot_dataset['atr'],
label='ATR (Average True Range)',
linewidth=linewidth)
plt.plot(plot_dataset['trix'],
label='TRIX, default to 12 days',
linewidth=linewidth)
plt.plot(plot_dataset['trix_9_sma'],
label='MATRIX is the simple moving average of TRIX',
linewidth=linewidth)
plt.legend()
plt.subplot(2, 1, 2)
plt.plot(plot_dataset['cci'],
label='CCI, default to 14 days',
linestyle='-',
linewidth=linewidth,
color='r')
plt.plot(plot_dataset['cci_20'],
label='20 days CCI',
linestyle='-',
linewidth=linewidth,
color='g')
plt.bar(x, plot_dataset['vr'], label='VR, default to 26 days')
plt.bar(x,
-plot_dataset['vr_6_sma'],
label='MAVR is the simple moving average of VR')
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\35_cci_tr_vr.png')
plt.show()
# 6.DMI
plt.figure(figsize=(16, 10), dpi=150)
linewidth = 1
plt.subplot(3, 1, 1)
plt.title('DMI and DMA')
plt.bar(x,
plot_dataset['pdi'],
label='+DI, default to 14 days',
color='r')
plt.bar(x,
-plot_dataset['mdi'],
label='-DI, default to 14 days',
color='g')
plt.legend()
plt.subplot(3, 1, 2)
plt.plot(plot_dataset['dma'],
label='DMA, difference of 10 and 50 moving average',
linewidth=linewidth,
color='k')
plt.legend()
plt.subplot(3, 1, 3)
plt.plot(plot_dataset['dx'],
label='DX, default to 14 days of +DI and -DI',
linewidth=linewidth)
plt.plot(plot_dataset['adx'],
label='6 days SMA of DX',
linewidth=linewidth)
plt.plot(plot_dataset['adxr'],
label='ADXR, 6 days SMA of ADX',
linewidth=linewidth)
plt.legend()
if save:
plt.savefig('project\\feature_engineering\\36_close_DMI.png')
plt.show()
return dataset_tech
