–a
Less data would generally lead us to using model based methods. Model bases methods tend to be advantagous to data driven methods when the series to forecast is short. A possible exception would be that data driven naive methods could be effective with very short time series.
–b
It is likely that customers of a nearby hospital would exhibit similar behavior to patients of this medical clinic. If the nearby hospital had more extensive records, their data could be useful to extact expansive insight involving customer behavior for the medical clinic.
–c
An advantage of this approach is that it accounts for the (weekly) seasonality, does not requre a long historical series. Another advantage is that the naive forcast is data driven, so it does not need to adhere to parametric assumptions. Simplicity
A disadvantage of this approach is that it is not likley to be the best method to account for trends.
–d The (seasonal) naive approach take here is easily automated. The analyst or computer must simply look back to the last " matching day of the week" to produce the forecast.
–e
To improve the current approach, you could create an ensemble forecast that combined the seasonal naive (looking back to the last “day of the week, e.g. Monday”) and the simple naive (looking back to the last business day). In this case you could give 50% weight to the seasonal naive and 50% weight to the naive.
You could also blend the seasonal naive with the mean of the entire series along with the naive forecast to produce an alternative ensemble forecast. In this case you could give 33% weight to the seaonal naive, 33% weight to the naive and 34% weight to the mean of the entire series.
This approach is data driven, and uses extrapolation.
Is nice for short term forcast, and requires minimal data. Downside is that it fails to capture long and medium term phenomenon.
Model driven and causality.
Statistical approaches are model based and can use any combination of causal modeling and correlation.
Thy hybrid approach can be a combination on model based and data driven.
Combining forecasts often helps accuracy by evening out biases, as well as possibly adding attitional information and accounting for different aspects of reality.
Possible disadvantages of a hybrid approach include increased costs and you would need analysts to be familar with multiple methods. You would also want to set the weighting of each forecast ahead of time to reduce bias.