I am implementing an ARIMA model for time series data. Since the data is not stationary I am performing data transformation log and performing exponential decay over the data.
Taking log of the data
passenger_log = np.log(indexdf['#Passengers'])
Then performing exponential decay of the log series
passenger_expdecay=passenger_log.ewm(halflife=12, min_periods=0, adjust=True).mean()
plt.plot(passenger_log)
plt.plot(passenger_expdecay, color='red')
The ADCF test shows better results for the exponential decay series (passenger_expdecay) compared to log series (passenger_log).
I want to use the exponential series as an input to ARIMA model but I dont know how to perform the inverse of this ewm function so that after prediction I can perform inverse transformation to get original values.
Can anybody help to perform inverse transformation of the exponential weighted (ewm) function
if you apply this as
I think this might Help...