**Measuring Prediction
Validity for Exponential Smoothing**

Introduction.
Now we need to focus on trying to get some sense of validity for our
prediction. In many respects this entire notion is ludicrous, but
that has failed to stop a lot of efforts. Sit back and think about
our approach.
This will be the basis of our validation
efforts. We will focus on
While there is no perfect measurement of the validity of a prediction, this should help us to compare validity to other forecasts. Now we need to develop our spreadsheet to illustrate this development. The formula for cell D4 should be
Then you can copy this down the column where comparisons exist. |

The formula for cell C13 where you take the average of
the percentage errors is
Notice that are percentage errors for each prediction are quite bad. At this point, all I can say is hopefully we will do better. We need to take the absolute value of each difference to make sure we average error terms which should all be positive.
Notice that we have used a different function to compute the errors. You should also notice that we had to recalculate the predictions in order to determine the errors. This code will be discussed in class. Make sure you test this code by comparing it to your results for the spreadsheet calculations. You are likely to want to develop your spreadsheet like the following so that it automatically recalculates both the forecast and the error as you change the weight. |