def get_daily_legs(self, jointi, open, high, low, close, volume):
avg05 = lines.get_ma(close, 5)
avg25 = lines.get_ma(close, 25)
o = open[jointi-WATCH_DAYS+1:jointi+1]
h = high[jointi-WATCH_DAYS+1:jointi+1]
l = low[jointi-WATCH_DAYS+1:jointi+1]
c = close[jointi-WATCH_DAYS+1:jointi+1]
v = volume[jointi-WATCH_DAYS+1:jointi+1]
ot = olines=[avg05[jointi-WATCH_DAYS+1:jointi+1], avg25[jointi-WATCH_DAYS+1:jointi+1]]
return (o, h, l, c, v, ot)
def get_line_data(self):
xbatch = []
ybatch = []
data = self.data
meigaras = data.keys()
random.shuffle(meigaras)
for code in meigaras[:BATCH_SIZE]:
(indexes, dates, open, high, low, close, volume) = data[code]
(pnf_joints, pnf_indexes, pnf_pmmask) = lines.get_pnf(high, low, close)
avg05 = lines.get_ma(close, 5)
avg25 = lines.get_ma(close, 25)
(pnf_indexes, pnf_pmmask) = self.add_mid_joints(pnf_indexes, pnf_pmmask)
idxs = range(1, len(pnf_indexes))
random.shuffle(idxs)
for i in idxs[:BATCH_SIZE]:
jointi = pnf_indexes[i]
if jointi >= INVALID_DAYS+WATCH_DAYS:
ylist = [0, 1, 0]
if pnf_pmmask[i] == -1:
ylist = [1, 0, 0]
if pnf_pmmask[i] == 1:
ylist = [0, 0, 1]
if self.mode == "DAILY":
(o, h, l, c, v, ot) = self.get_daily_legs(jointi, open, high, low, close, volume)
elif self.mode == "WEEKLY":
(o, h, l, c, v) = self.get_monthly_legs(jointi, open, high, low, close, volume)
ot = []
else:
raise Exception("mode is not correctly set")
xlist = self.lines2image(o, h, l, c, v, ot)
ylist = np.array(ylist)
if len(xbatch) == 0:
xbatch = np.array([xlist])
ybatch = np.array([ylist])
else:
xbatch = np.append(xbatch, [xlist], axis=0)
ybatch = np.append(ybatch, [ylist], axis=0)
return [xbatch, ybatch]
def get_xs(self, code, data, starti, endi):
#(indexes, dates, open, high, low, close, volume) = kf.get_kabuka([code], "", endd, True)
(indexes, dates, open, high, low, close, volume) = data
avg = lines.get_ma(close)
xs = []
for i in range(starti, endi+1):
ximage = self.lines2image(open[-WATCH_DAYS+i:i],
high[-WATCH_DAYS+i:i],
low[-WATCH_DAYS+i:i],
close[-WATCH_DAYS+i:i],
[avg[-WATCH_DAYS+i:i]])
xs.append(ximage)
return np.array(xs)
def get_line_data(self, meigaras=[]):
data = kf.get_kabuka(meigaras, self.startd, self.endd, True)
xbatch = []
ybatch = []
for code in data.keys():
(indexes, dates, open, high, low, close, volume) = data[code]
(pnf_joints, pnf_indexes, pnf_pmmask) = lines.get_pnf(high, low, close)
avg = lines.get_ma(close)
(pnf_indexes, pnf_pmmask) = self.add_mid_joints(pnf_indexes, pnf_pmmask)
for i in range(1, len(pnf_indexes)):
jointi = pnf_indexes[i]
if jointi >= MA_DAYS+WATCH_DAYS:
#xlist = open[jointi-5+1:jointi+1]
#xlist += high[jointi-5+1:jointi+1]
#xlist += low[jointi-5+1:jointi+1]
#xlist += close[jointi-5+1:jointi+1]
ylist = [0, 1, 0]
if pnf_pmmask[i] == -1:
ylist = [1, 0, 0]
if pnf_pmmask[i] == 1:
ylist = [0, 0, 1]
#xlist = self.normalize(np.array(xlist))
#xlist = np.append(xlist, self.normalize_volume(volume[jointi-5+1:jointi+1]))
#xlist = np.append(xlist, np.zeros(XSIZE-len(xlist)))
xlist = self.lines2image(
open[jointi-WATCH_DAYS+1:jointi+1],
high[jointi-WATCH_DAYS+1:jointi+1],
low[jointi-WATCH_DAYS+1:jointi+1],
close[jointi-WATCH_DAYS+1:jointi+1],
olines=[avg[jointi-WATCH_DAYS+1:jointi+1]])
ylist = np.array(ylist)
if len(xbatch) == 0:
xbatch = np.array([xlist])
ybatch = np.array([ylist])
else:
xbatch = np.append(xbatch, [xlist], axis=0)
ybatch = np.append(ybatch, [ylist], axis=0)
return [xbatch, ybatch]
def get_line_data(self):
xbatch = []
ybatch = []
data = self.data
meigaras = data.keys()
random.shuffle(meigaras)
for code in meigaras[:BATCH_SIZE]:
(indexes, dates, open, high, low, close, volume) = data[code]
(pnf_joints, pnf_indexes, pnf_pmmask) = lines.get_pnf(high, low, close)
avg05 = lines.get_ma(close, 5)
#avg25 = lines.get_ma(close, 25)
(pnf_indexes, pnf_pmmask) = self.add_mid_joints(pnf_indexes, pnf_pmmask)
idxs = range(1, len(pnf_indexes))
random.shuffle(idxs)
for i in idxs[:BATCH_SIZE]:
jointi = pnf_indexes[i]
if jointi >= INVALID_DAYS+WATCH_DAYS:
if i >= len(pnf_indexes)-1:
continue
ylist = [0, 1, 0]
if pnf_indexes[i+1]-pnf_indexes[i]<MIN_LEG_LEN:
ylist = [0, 0, 0]
else:
if pnf_pmmask[i] == -1:
ylist = [1, 0, 0]
if pnf_pmmask[i] == 1:
ylist = [0, 0, 1]
(o, h, l, c, v, ot) = self.get_daily_legs(jointi, open, high, low, close, volume)
xlist = self.lines2image(o, h, l, c, v, ot)
ylist = np.array(ylist)
if len(xbatch) == 0:
xbatch = np.array([xlist])
ybatch = np.array([ylist])
else:
xbatch = np.append(xbatch, [xlist], axis=0)
ybatch = np.append(ybatch, [ylist], axis=0)
return [xbatch, ybatch]
