This is a great example. You can take either a computational approach or a learning approach to forecasting.
With a learning approach, you can train a forecasting model on the past data. This can be as simple as a linear regression where you are fitting a curve to the dataset. This is good if you expect the future to follow similar patterns to the past. But, this is often not the case and a regression model can’t factor in those new circumstances because they aren’t represented in the training data. As an alternative, you can take a computational approach where you model a set of assumptions and make the prediction based on those assumptions.
This is a great example of how both approaches are valid depending on the circumstances.
Is forecasting computation problem or learning problem?
This is a great example. You can take either a computational approach or a learning approach to forecasting.
With a learning approach, you can train a forecasting model on the past data. This can be as simple as a linear regression where you are fitting a curve to the dataset. This is good if you expect the future to follow similar patterns to the past. But, this is often not the case and a regression model can’t factor in those new circumstances because they aren’t represented in the training data. As an alternative, you can take a computational approach where you model a set of assumptions and make the prediction based on those assumptions.
This is a great example of how both approaches are valid depending on the circumstances.
Thank you🙏