Eikon Data API with Time Series Forecasting

    	
            
get_timeseries(rics, fields='*', start_date=None, end_date=None, interval='daily', count=None, calendar=None, corax=None, normalize=False, raw_output=False, debug=False)

        
        Copy
    

    	
            
import pandas as pd
import eikon as ek
 
#set app key to identify the application on Refinitiv Platform
ek.set_app_key(<APP_KEY>)
#a RIC that you want to get time series
RIC = 'IBM.N'
#start date that you want to get time series; it is in format yyyy-mm-dd
start_date = '2018-01-02'
#get time series till the last market close date using eikon.get_timeseries(..) function
df = ek.get_timeseries(RIC, 'CLOSE', interval='daily', start_date=start_date)
#display time series data frame
df

        
        Copy
    

    	
            
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
 
from datetime import timedelta, datetime
from statsmodels.tsa.arima_model import ARIMA
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from statsmodels.tsa.stattools import acf, pacf
from statsmodels.tsa.stattools import adfuller
 
#function to forecast using ARIMA Model
#input is time series data frame(df) got from eikon.get_timeseries(..)
#and forecast end date which is in format yyyy-mm-dd
def ARIMA_model_forecast(df, forecast_end_date):
    #Set the start and end forecast date time
    #first date is 1 day after time series. The last date is the forecast_end_date parameter
    _e_date = datetime.fromtimestamp(datetime.timestamp(df.index[-1]))
    e_date = _e_date + timedelta(days = 1)
    e_date = e_date.strftime('%Y-%m-%d')
 
    #find differences of time series which is input of ARIMA model
    df['df_1diff'] = df['CLOSE'].diff()
    df = df[df['df_1diff'].notnull()] 
    lag_acf = acf(df['df_1diff'], nlags=50)
    lag_pacf = pacf(df['df_1diff'], nlags=50, method='ols')
    new_spot = df['df_1diff'].resample('D').ffill()
    #call ARIMA model which p=1,d=0 and q=1 with differences of time series  
    arma_model = ARIMA(new_spot, order=(1,0,1)) 
    results = arma_model.fit(disp=0)
    residuals = pd.DataFrame(results.resid)
    #focast time series
    predictions_ARIMA = pd.Series(results.fittedvalues, copy=True)
    predictions_ARIMA_cumsum = predictions_ARIMA.cumsum()
    predictions_ARIMA_final = pd.Series(df['CLOSE'].iloc[0], index=new_spot.index)
    predictions_ARIMA_final = predictions_ARIMA_final.add(predictions_ARIMA_cumsum, fill_value=0)
    new_spot = df.CLOSE.resample('D',label='right').ffill()
    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=forecast_end_date)
 
    #create data frame from forecast timeseries  
    preds = pd.DataFrame(index= pd.date_range(start=e_date, end=forecast_end_date), data=predicted_values, columns=['CLOSE'])
 
    #Plot graph of past and forecast timeseries
    plt.figure(figsize=(16,7))
    plt.plot(new_spot, label='Actual')
    plt.plot(preds, label='Forecast', color='red')
    plt.legend(loc='best')
    plt.show()
    return preds
 
#end date that you want to forecast
forecast_end_date = '2020-03-31'
#call the function to forecast time series with ARIMA model
forecast = ARIMA_model_forecast(df,forecast_end_date)
#display forecast time series
forecast

        
        Copy
    

    	
            
Error on checking port 9000 : HTTPConnectionPool(host='localhost', port=9000): Max retries exceeded with url: /api/v1/data (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x000001C4C33D47C8>: Failed to establish a new connection: [WinError 10061] No connection could be made because the target machine actively refused it'))
...

        
        Copy
    

    	
            
2020-02-20 10:22:22,228 P[22984] [MainThread 20416] Error code 403 | Client Error: {"ErrorCode":1401,"ErrorMessage":"Application id <APP_KEY> is invalid"}

        
        Copy