Você está na página 1de 1

FORECASTING METHODS FOR MANAGERS

TUTORIAL 4 – Optimisation and Out of Sample Fitting


The following data representing the sales of washing powder at a local supermarket.

Week 1 2 3 4 5 6 7 8
Sales (boxes) 124 123 124 121 119 119 125 117

In this tutorial we will use the forecasts and error calculations which you produced over the past 2 weeks.
For each model type (MA, WMA and SES) we will find the parameter values which produce the most
accurate model.

We will then look at a refined model construction and selection method which is often better at finding
the model which is likely to predict the future most accurately.

Tutorial Tasks

1. For the SES model, use Excel’s Solver feature to find the value of alpha which gives the best
MAE, MAPE, MSE and RMSE. Solver can be loaded using the same process which you may
have used to load the ‘Data Analysis’ Add-in.

2. Use Solver to find the 3 weightings for the WMA3 model which minimise the MSE.

3. What value of ‘N’ for the MA models gives the lowest MSE?

4. For these 3 model types (MA1 to MA4, SES, WMA3) which is the best overall model - based on
the lowest MSE?

Forecasting Methods for Managers 1 Tutorial 4

Você também pode gostar