Hello,
I am working om forecasting task of sales daily data with high frequency on weekdays and low on weekends, but lower mostly on saturdays. I am modeling weekly seasonality period 7, fourier hypertuning( 3 to 10), yearly seasonality period 365.25, fourier (10 to 25), i also added holidays prior scale , changepoint range, seasonality prior scale for the weekly and yearly and added regressor for saturday and regressor for sunday. I notice that predictions on holidays and weekends that are holidays (real values really low ), predictions are huge. Also weekend (mostly saturdays) on peak season such december, november, january are predicted larger than usual. I do not know how to treat seasonalities, different for weekdays and different for weekends though most different is saturdays. If i add another regressor saturday for december ,should i remove saturdays from the other regressor is_saturday.d Does the weekly seasonality conflicts with separated regressors of saturday and sunday? What about christmas, black friday ,if i model as separated weeks custom seasonalities, should i also model normal weeks non blacks friday, normal weeks non christmas custom seasonalities? What would be your suggestions in such case? I tried ,but i am confused..
Hello, I am working om forecasting task of sales daily data with high frequency on weekdays and low on weekends, but lower mostly on saturdays. I am modeling weekly seasonality period 7, fourier hypertuning( 3 to 10), yearly seasonality period 365.25, fourier (10 to 25), i also added holidays prior scale , changepoint range, seasonality prior scale for the weekly and yearly and added regressor for saturday and regressor for sunday. I notice that predictions on holidays and weekends that are holidays (real values really low ), predictions are huge. Also weekend (mostly saturdays) on peak season such december, november, january are predicted larger than usual. I do not know how to treat seasonalities, different for weekdays and different for weekends though most different is saturdays. If i add another regressor saturday for december ,should i remove saturdays from the other regressor is_saturday.d Does the weekly seasonality conflicts with separated regressors of saturday and sunday? What about christmas, black friday ,if i model as separated weeks custom seasonalities, should i also model normal weeks non blacks friday, normal weeks non christmas custom seasonalities? What would be your suggestions in such case? I tried ,but i am confused..