I'm currently doing some forecasting on daily data and I was wondering whether is there a problem whenever i add exogenous variables (like the day of the week or month of the year) provided as one-hot encoded vectors? Isn't there a risk of reaching the known "dummy trap problem", as it is known to occur in traditional regression models and SARIMA? If this is true, then I would like to know whether the prophet library takes this into account when including such types of regressors.
Prophet already models day-of-week and day-of-year seasonality using Fourier series as extra regressors; it's not recommended to add additional dummy variables for these.
I'm currently doing some forecasting on daily data and I was wondering whether is there a problem whenever i add exogenous variables (like the day of the week or month of the year) provided as one-hot encoded vectors? Isn't there a risk of reaching the known "dummy trap problem", as it is known to occur in traditional regression models and SARIMA? If this is true, then I would like to know whether the prophet library takes this into account when including such types of regressors.
Thank you