def calculate_rsi(date, coin, screen):
tickers_df = tickers.get_tickers(15, coin, date, screen)
if len(tickers_df) >= 15:
profit = []
loss = []
df = tickers_df.iloc[(len(tickers_df) - 15):len(tickers_df)]
for i in range(len(df)):
if i == 0:
continue
last_d = df.price.values[i]
last_d1 = df.price.values[(i - 1)]
dif = last_d - last_d1
if dif >= 0:
profit.append(dif)
else:
loss.append(dif * (-1))
if len(profit) == 0:
rs = 0
else:
if len(loss) == 0:
rs = 100
else:
rs = np.mean(profit) / np.mean(loss)
# RSI = 100 - 100 / (1 + RS)
rsi_value = 100 - (100 / (1 + rs))
rsis.insert_rsi(date, coin, rsi_value, screen)
def calculate_macd(date, coin, screen):
signal_line = np.NaN
histogram = np.NaN
tickers_df = tickers.get_tickers(26, coin, date, screen)
if len(tickers_df) >= 26:
ema_12 = tickers_df.iloc[len(tickers_df) - 12:len(tickers_df)].drop(
['date', 'coin'], axis=1).price.ewm(span=12,
min_periods=12,
adjust=True,
ignore_na=False).mean()
ema_26 = tickers_df.iloc[len(tickers_df) - 26:len(tickers_df)].drop(
['date', 'coin'], axis=1).price.ewm(span=26,
min_periods=26,
adjust=True,
ignore_na=False).mean()
ema12 = ema_12.iloc[11]
ema26 = ema_26.iloc[25]
macd_line = ema_12 - ema_26
macd_df = macds.get_macds(9, coin, date, screen)
if len(macd_df) >= 9:
signal_line = macd_df.iloc[len(macd_df) -
9:len(macd_df)].macd_line.ewm(
span=9,
min_periods=9,
adjust=True,
ignore_na=False).mean()
signal_line = signal_line[0]
histogram = macd_line[0] - signal_line
macds.insert_macd(date, coin, ema12, ema26, macd_line[0], signal_line,
histogram, screen)
def triple_screen(n, coin, screen):
tickers_df = tickers.get_tickers(n, coin, datetime.now(), 0)
if len(tickers_df) >= n:
# last = tickers_df.price.mean()
last = tickers_df.iloc[0].price
tickers.insert_tickers(datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
coin, last, screen)
def calculate_sma(date, coin, screen):
tickers_df = tickers.get_tickers(20, coin, date, screen)
if len(tickers_df) >= 5:
sma5 = tickers_df.iloc[(len(tickers_df) -
5):len(tickers_df)].price.mean()
sma5_theta = calculate_theta(date, coin, sma5, screen)
if len(tickers_df) >= 20:
sma20 = tickers_df.iloc[(len(tickers_df) -
20):len(tickers_df)].price.mean()
smas.insert_smas(date, coin, sma5, sma20, sma5_theta, screen)
def get_boillinger_vote(date, coin, screen):
vote = 0
boillinger_df = boillingers.get_boillinger(1, coin, date, screen)
if not boillinger_df.empty:
tickers_df = tickers.get_tickers(1, coin, date)
if not tickers_df.empty:
price = tickers_df.iloc[0].price
lower_band = boillinger_df.iloc[0].lower_band
upper_band = boillinger_df.iloc[0].upper_band
if price < lower_band:
vote = 1
elif price > upper_band:
vote = -1
return vote
def calculate_boillinger(date, coin, screen):
tickers_df = tickers.get_tickers(20, coin, date, screen)
if len(tickers_df) >= 20:
sma20 = tickers_df.iloc[(len(tickers_df) -
20):len(tickers_df)].price.mean()
std20 = tickers_df.iloc[(len(tickers_df) -
20):len(tickers_df)].price.std()
ema20 = emas.get_emas(1, coin, date, screen).iloc[0].ema20
# upper_band = sma20 + 2 * std20
# lower_band = sma20 - 2 * std20]
#TODO: CHANGE PERCENTAGE TO A LIL BIT BIGGER
upper_band = ema20 * (1 + 0.15)
lower_band = ema20 * (1 - 0.15)
height = upper_band - lower_band
boillingers.insert_boillingers(date, coin, upper_band, lower_band,
ema20, height, screen)
def calculate_ema(date, coin, screen):
tickers_df = tickers.get_tickers(20, coin, date, screen)
if len(tickers_df) >= 5:
ema5 = tickers_df.iloc[(len(tickers_df) -
5):len(tickers_df)].price.ewm(
span=5,
min_periods=5,
adjust=True,
ignore_na=False).mean()
ema5 = ema5.iloc[4]
if len(tickers_df) >= 20:
ema20 = tickers_df.iloc[(len(tickers_df) -
20):len(tickers_df)].price.ewm(
span=20,
min_periods=20,
adjust=True,
ignore_na=False).mean()
ema20 = ema20.iloc[19]
emas.insert_emas(date, coin, ema5, ema20, screen)
def trend_market(date, coin, ema_dif):
from scipy.fftpack import ifft, fft
from math import atan
# 300 segundos = 5 min
# 300 * 4h = 1200 h / 24 h = 50 dias
#
df = tickers.get_tickers(300, coin, date, 1)#TODO Change screen test
df = df[df.date < date]
data = df.price
if len(df) < 4:
vote = 0
return
else:
get_max_min(coin, df)
yf = fft(data)
wn = 18
yf[wn:-wn] = 0
iY = ifft(yf).real[::-1]
theta = [atan((iY[0] - iY[1]) / 48),
atan((iY[1] - iY[2]) / 48)]
d_theta = (theta[0] - theta[1]) / theta[1]
n = 2
macd_df_one = macds.get_macds(n, coin, date, 1)
if len(macd_df_one) < n:
return None
vote = 0
current_ema26 = macd_df_one.iloc[1].ema_26
current_ema12 = macd_df_one.iloc[1].ema12
dif_current = np.log(current_ema12/current_ema26)
base_ema26 = macd_df_one.iloc[0].ema_26
base_ema12 = macd_df_one.iloc[0].ema12
dif_base = np.log(base_ema12/base_ema26)
delta_dif = (dif_current - dif_base)/dif_base
vote_rsi = rsi_sign(coin, date)
if vote_rsi is None:
return None
theta_b = features.features_signal_theta(theta[1])
theta_c = features.features_signal_theta(theta[0])
dif_b = features.features_signal_dif(dif_base)
dif_c = features.features_signal_dif(dif_current)
# if theta[0] > 0 and dif_current > 0:
# vote = 1
# else:
# vote = 0
filename = '../Notebooks/finalized_model.sav'
loaded_model = pickle.load(open(filename, 'rb'))
[dif_current, theta[0], rsi_sign(coin, date), ]
result = loaded_model.predict(X_test, Y_test)
print(result)
insert_trend(coin, date, 1, dif_current, dif_base, delta_dif, theta[0], theta[1], d_theta, vote)
return vote