Do You Have to Deal With Noise or Imprecision in Forecasting?

A lot of times there are outliers in the form of misleading aggregations. If you don't collect information well, you can sometimes have tripler times the normal car activity. The solonoid doesn't know if it's a holiday, for example, or if there's a sporting event. So i definitely see how that signal could be useful in making future predictions. But we don't want to smooth to match the signal, because you also lose quality information.