Forecasting Flashcards
Time Series Forecasting
Plot data to identify any patterns
Identify which components present
Decide on appropriate model
Time Series Models
Additive A = T + S + E
Multiplicative A = T x S x E
Patterns Analysis
Each week plotted as a separate line shows a similar pattern for each week
Additive most appropriate they a similar nature
Multiplicative much more erratic
Building Future Forecast
Estimated Trend
Seasonal Component
Reliability of Forecast
Does the forecast fit the data?
Historical Data
Assess residuals
Forecasting error
Decomposition Procedure
A = T + S + E - Average to eliminate S leave T + some E
Deduct T from A to leave S and some E: A - T = S + some E
Refine S
Use S to re-estimate T and build model for T: A - S = T + E
Estimate residual A - T - S + = E