from statsmodels.tsa.stattools import adfuller
from statsmodels.tsa.stattools import acf, pacf
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.tsa.seasonal import seasonal_decompose
from datetime import timedelta, datetime
import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
import refinitiv.data as rd
import warnings
warnings.filterwarnings('ignore')
plt.style.use('dark_background')
rd.open_session()
def get_data(ric, s_date):
spot_df = rd.get_history(
ric, 'TRDPRC_1', interval='daily', start=s_date, count=10000)
_e_date = datetime.fromtimestamp(datetime.timestamp(spot_df.index[-1]))
e_date = _e_date + timedelta(days=1)
e_date = e_date.strftime('%Y-%m-%d')
_e_date = _e_date + timedelta(days=30)
fce_date = _e_date.strftime('%Y-%m-%d')
decomposition = seasonal_decompose(spot_df.TRDPRC_1, period=12)
trend = decomposition.trend
seasonal = decomposition.seasonal
residual = decomposition.resid
spot_df['spot_1diff'] = spot_df['TRDPRC_1'].diff()
spot_df = spot_df[spot_df['spot_1diff'].notnull()] # drop null rows
lag_acf = acf(spot_df['spot_1diff'], nlags=50)
lag_pacf = pacf(spot_df['spot_1diff'], nlags=50, method='ols')
new_spot = spot_df['spot_1diff'].resample(
'D').ffill() # resample per day and fill the gaps
new_spot = new_spot.bfill()
new_spot = new_spot.astype('float')
arma_model = ARIMA(new_spot, order=(1, 0, 1))
results = arma_model.fit()
residuals = pd.DataFrame(results.resid)
predictions_ARIMA = pd.Series(results.fittedvalues, copy=True)
predictions_ARIMA_cumsum = predictions_ARIMA.cumsum()
predictions_ARIMA_final = pd.Series(
spot_df['TRDPRC_1'].iloc[0], index=new_spot.index)
predictions_ARIMA_final = predictions_ARIMA_final.add(
predictions_ARIMA_cumsum, fill_value=0)
new_spot = spot_df.TRDPRC_1.resample(
'D', label='right').ffill().astype('float')
es_model = ExponentialSmoothing(
new_spot, trend='add', damped=False, seasonal='mul', seasonal_periods=30)
es_results = es_model.fit()
predicted_values = es_model.predict(
params=es_results.params, start=e_date, end=fce_date)
preds = pd.DataFrame(index=pd.date_range(
start=e_date, end=fce_date), data=predicted_values, columns=['TRDPRC_1'])
plt.figure(figsize=(16, 7))
plt.plot(new_spot, label='Actual')
plt.plot(preds, label='Forecast', color='pink')
plt.legend(loc='best')
plt.show()
return preds
