Speed up auto_arima on time series with seasonality of 2 years

问题

I have a time series at weekly level with seasonlity of 2 years. I am using auto_arima to forecast for next 2 years. Plot of the time series -

The challenge is - its taking more than 90 mins to train the model with following parameters.

stepwise_model = auto_arima(onion_data_train_weekly['Modal_Price'], start_p=1, start_q=1,
                           max_p=1, max_q=1, m=104,
                           start_P=0, seasonal=True,
                           d=1, D=1, trace=True,
                           error_action='ignore',  
                           suppress_warnings=True, 
                           stepwise=True)
print(stepwise_model.aic())

Note: Weekly data with seasonlity of 2 years translates to m value of 52*2 =104. As a fact I know that greater the value of m slower is the ARIMA model to train.

Any suggestions to speed it up or alternative options?

来源：https://stackoverflow.com/questions/60249902/speed-up-auto-arima-on-time-series-with-seasonality-of-2-years