HP 3000 Manuals

HP RXForecast Users Manual for MPE Systems

Exponential Example 

Suppose you generate the following Stats report from an exponential
forecast on the Prompt to Response time metric:


            

Notice the formula used to calculate this forecast:

FC = EXP (INT + SLOPE * TIME + DOW + WOM + MOQ + QOY)

The argument within parentheses is the same as the linear calculation:
you need to know the intercept, slope, and the number of time periods
since the start of the forecast, as well as seasonality.  Once you have
calculated the parenthetic, you calculate EXP, or e, raised to that
number.

Question 

What is the forecast for the average Response to Prompt times (in
seconds) for 1 June 1988?

Answer 

Calculate the statement within parentheses.  Then apply the EXP
function--that is, raise e to this power, where e is the constant 2.718.

Most of the numbers you need (intercept, slope, seasonality) are given in
the Stats report.  Use a calendar to calculate TIME. Insert these numbers
into the formula as follows:

Forecast  = EXP (0.036496 + 0.0012084 * 92)

          = EXP (0.036496 + 0.1111728)

          = e0.1476688 

          = 1.1591289

Prompt to Response times on 1 June 1988 are forecasted to be 1.16
seconds.

These numbers were derived as follows:

   1.  Calculate TIME, the number of days from the start date (1 March
       1988) through 1 June.

       This number is inclusive, so it is 13 weeks plus 1 day, or 92
       days.

   2.  Multiply TIME, 92, by the SLOPE (0.0012084).

       Note that 1.2084e-003 is read as 1.2084 raised to the 10-3 power.
       Simply move the decimal point to the left 3 places.

   3.  Add the INTERCEPT (0.036496).

       Again, the intercept is 3.6496e-002.  So move the decimal point to
       the left 2 places.

   4.  There is no seasonality to add.

   5.  Then raise EXP, or e, to the number calculated in the first three
       steps above.

The figure below shows the Stats report that is generated from another
Prompt to Response forecast.  The only difference between this forecast
and the one displayed earlier is that Ignore Weekends is ON.


            

Question 

What is the forecast for the average Response to Prompt times (in
seconds) for 1 June 1988?

Answer 

Insert the numbers into the formula as follows:

Forecast  = EXP (0.084194 - 0.0059054 * 66)

          = EXP (0.084194 - 0.3897564)

          = e-0.3055624 

          = .73670893

Prompt to Response times on 1 June 1988 are forecasted to be .737
seconds.

These numbers were derived as follows:

   1.  Calculate TIME, the number of days from the start date (1 March
       1988) through 1 June.

       This number is inclusive, so it is 13 weeks plus 1 day, or 92 days
       less 13 weekends.  Thus TIME is 66.

       _________________________________________________________________ 

       NOTE  Since Ignore Weekends is ON, weekend days are excluded from
             the TIME count.

       _________________________________________________________________ 

   2.  Multiply TIME, 66, by the SLOPE (-0.0059054).

   3.  Add the INTERCEPT (0.084194).

   4.  There is no seasonality to add.

   5.  Then raise EXP, or e, to the number calculated in the first three
       steps above.

MPE/iX 5.0 Documentation