6 Business Forecasting Free Online FlashCards

What distinguishes a forecast from a goal or plan?

A forecast estimates an unknown future value using information available when it is made; unlike a goal or plan, it is not a desired outcome or a set of actions.

What does a prediction interval communicate?

A prediction interval gives a range of plausible future outcomes. It helps express uncertainty that a single point forecast cannot show.

How does seasonality differ from a cycle?

Seasonality repeats at a known, fixed interval; a cycle is a longer-term fluctuation without a fixed, reliably repeating period.

What is a baseline forecast used for?

A baseline is a simple benchmark forecast. A more elaborate method is useful only if it performs better on data not used to construct it.

How does the mean method forecast future values?

The mean method forecasts every future period using the historical average: $\hat{y}_{T+h|T}=\frac{y_1+\cdots+y_T}{T}$. It can suit a series fluctuating around a stable level.

How is forecast error calculated, and what does its sign mean?

Forecast error is actual minus forecast: $e_t=y_t-\hat{y}_t$. A positive error means the actual value exceeded the forecast; a negative error means the forecast was too high.

How do forecast errors differ from residuals?

Forecast errors are measured on observations not used to make the forecasts. Residuals are calculated on data used to fit a model, so small residuals alone do not prove good forecasting.

What does MAE measure?

Mean absolute error is the average absolute forecast error: $\operatorname{MAE}=\operatorname{mean}(|e_t|)$. It reports a typical miss in the series’ original units; lower is better.

What is the naïve forecast for each future period?

The naïve method sets every future forecast equal to the latest observation: $\hat{y}_{T+h|T}=y_T$.

How does the seasonal naïve method forecast a period?

It uses the latest observed value from the matching season. For example, a future March forecast uses the most recent observed March.

What does the drift method extend?

The drift method extends the average change between the first and last observations: $\hat{y}_{T+h|T}=y_T+h\frac{y_T-y_1}{T-1}$. It assumes that average change continues.

How should time-series forecasts be tested?

Train on earlier observations and test against later ones; do not shuffle time-series data. Rolling-origin evaluation repeats this process at several forecast origins.