The FCSET command specifies the characteristics that you want the Geneva Forecasting engine to use when executing a forecasting context created using a FCOPEN statement.
You must use a FCSET statement in combination with other OLAP DML statements as outlined in "Forecasting Programs".
FCSET handle-expression forecast-characteristic
where forecast-characteristic has the following syntax:
An INTEGER
expression that is the handle to forecast context that you want to query and that was previously opened using the FCOPEN function.
Indicates whether the forecast engine reduces the risk of over-adjustment by allocating or forecasting the last cycle.
NO specifies that the forecast engine forecasts the last cycle. (Default)
YES specifies that the forecast engine forecasts only the average value for one period of the cycle. That average value is then multiplied by factors to give the remaining points in that period. For example, when the last cycle has 24-hour periods, only an average hourly value is forecast, which is then multiplied by 24 hourly factors to give the value for each hour.
For the single exponential smoothing, double exponential smoothing, and Holt-Winters forecasting methods, specifies the value for Alpha which is the baseline parameter that is used for those methods.
ALPHA MAX decimal, specifies the maximum value of Alpha. For decimal, you can specify any decimal value from 0.0 through 1.0. The default value of decimal is 0.3.
ALPHA MIN decimal specifies the minimum value of Alpha. For decimal, you can specify any decimal value from 0.0 through 1.0. The default value of decimal is 0.1.
ALPHA STEP decimal specifies the value of the interval that the forecasting engine uses when it determines the value of Alpha. For decimal, you can specify any decimal value from 0.05 through 0.2 as long as the value evenly divides the difference between the values of ALPHA MAX and ALPHA MIN. The default value of decimal is 0.1.
Specifies the approach that the forecasting engine takes when it executes the forecast.
'APPAUTO' indicates that the forecasting engine tests all of possible forecasting methods and options for these methods and chooses and uses the method that best fits the data. When you specify this value, the forecasting engine ignores any value that you specify for the METHOD keyword. (Default)
'APPMANUAL' indicates that the Geneva Forecasting engine uses the method and options you specify in this FCSET statement when executing the forecast.
For the double exponential smoothing and Holt-Winters forecasting methods, specifies the value of Beta. Beta is the trend parameter that controls the estimate of the trend.
BETA MAX decimal specifies the maximum value of Beta. For decimal, you can specify any decimal value from 0.0
through 1.0
. The default value is 0.3
.
BETA MIN decimal specifies the minimum value of Beta. For decimal, you can specify any decimal value from 0.0
through 1.0
. The default value is 0.1
.
BETA STEP decimal specifies the value of the interval that the forecasting engine uses when it determines the value of Beta. For decimal, you can specify any decimal value from 0.05
through 0.2
as long as the value evenly divides the difference between the values of BETA MAX and BETA MIN. The default value of decimal is 0.1
.
Indicates whether optimization should be done on the median smoothed data series.
NO specifies that the methods are done using the original historical time series data. (Default)
YES specifies that optimization is done on the median smoothed data series, which results in more smoothed or "baseline" forecasts.
For linear and nonlinear regression methods, specifies the value of the cyclical decay. Cyclical decay pertains to how seriously the forecasting engine considers deviations from baseline activity when it performs linear and nonlinear regressions.
CYCDECAY MAX decimal, specifies the maximum value of the cyclical decay parameter. For decimal, you can specify any decimal value from 0.2
through 1.0
as long as the difference between the values of CYCDECAY MIN and CYCDECAY MAX is evenly divided by 0.4
. The default value of decimal is 1.0
.
CYCDECAY MIN decimal, specifies the minimum value of the cyclical decay parameter. For decimal, you can specify any decimal value from 0.2 through 1.0 as long as the difference between the values of CYCDECAY MIN and CYCDECAY MAX is evenly divided by 0.4. The default value of decimal is 0.2.
For the Holt-Winters forecasting method, specifies the value of Gamma which is the seasonal parameter used by that method.
GAMMA MAX decimal specifies the maximum value of Gamma. For decimal, you can specify any decimal value from 0.0
through 1.0
. The default value of decimal is 0.3
.
GAMMA MIN decimal specifies the minimum value of Gamma. For decimal, you can specify any decimal value from 0.0
through 1.0
. The default value of decimal is 0.1
.
GAMMA STEP decimal specifies the value of the interval that the forecasting engine uses when it determines the value of Gamma. For decimal, you can specify any decimal value from 0.05
through 0.2
as long as the value evenly divides the difference between the values of GAMMA MAX and GAMMA MIN. The default value of decimal is 0.1
.
The number of historical periods. For integer
, you can specify any INTEGER
value from 1
through 50000
, which is the maximum number of time dimension values that can be present in the time-series expression specified in the FCEXEC command. (Note that the number of forecast periods is derived by subtracting the value of HISTPERIODS from the STATLEN of the dimension of the time-series expression.)
Specifies the upper bound on the forecast data. The number you specify for decimal indicates a multiple of the largest value in the historical series. For example, when you specify 10.0
, the upper bound will be 10 times the largest value in the historical series. The default value is 100.0
.
Specifies the forecasting method that you want the forecasting engine to use. Values that you specify for method are ignored unless the value of APPROACH
is set to 'APPMANUAL'
.
You can specify one of the following keywords for method:
AUTOMATIC specifies that the forecasting engine determines and uses the method that is the best fit for the data. (Default)
LINREG specifies the linear regression method in which a linear relationship (y=a*x+b)
is fitted to the data.
NLREG1 specifies a nonlinear regression method in which a linear relationship (y'=a*x'+b)
is fitted to a transformation of the original data; in this case, x'=log(x) and y'=log(y)
. This results in the development of a polynomial model between x
and y(y=c*x^a)
.
NLREG2 specifies a nonlinear regression method in which a linear relationship (y'=a*x'+b)
is fitted to a transformation of the original data; in this case, x'=x and y'=ln(y)
. This results in the development of an exponential model between x and y(y=c*e^ax)
.
NLREG3 specifies a nonlinear regression method in which a linear relationship (y'=a*x'+b)
is fitted to a transformation of the original data; in this case, x'=log(x)
and y'=y
. This results in the development of a logarithmic model between x and y(y=a*log(x)+b)
.
NLREG4 specifies a nonlinear regression method in which a linear relationship (y'=a*x'+b)
is fitted to a transformation of the original data; in this case, x'=1/x and y'=1/y
. This results in the development of an asymptotic curve (y=x/(a+bx))
.
NLREG5 specifies a nonlinear regression method in which a linear relationship (y'=a'*x+b)
is fitted to a transformation of the original data; in this case, x'=x and y'=ln(y/(K-y))
. This results in the development of an exponential asymptotic curve (y=cKe^ax/(1+ce^ax))
.
SESMOOTH specifies the single exponential smoothing method in which the current estimate is taken as a geometrically weighted average of past values, and all future values are given this same value. This method is intended for short term forecasts of non-seasonal data.
DESMOOTH specifies the double exponential smoothing method in which the current estimate is taken as a geometrically weighted average of past values, and this is added to a trend term calculated by the same method. Single exponential smoothing is therefore applied to both the series and the trend term.
CROSTON specifies the Croston's Intermittent Demand method. This is a forecasting method which is a variant of exponential smoothing that can be used for intermittent data (that is data where more than half of the observations are zero). This method first estimates the interval between positive demands, and then estimates the magnitude of the demand when positive.
HOLT/WINTERS specifies the Holt-Winters method that is used on seasonal data, in which double exponential smoothing methods with trend damping are combined with multiplicative seasonal factors, which are estimated using single exponential smoothing.
Specifies the lower bound on the forecast data. The number you specify indicates a multiple of the smallest value in the historical series. For decimal, you can specify any decimal value from 0.0
through 1.0
. For example, when you specify 0.5
the lower bound will be half the smallest value in the historical series. The default value of decimal is 0.0
.
Specifies the value of the parameter that the forecasting engine uses when it adjusts the decay of estimates of base values that it uses when it unravels the predictions on a moving periodic total (MPT) series.
MPTDECAY MAX decimal specifies the maximum value of the parameter that the forecasting engine uses when it adjusts the decay of estimates of base values that it uses when it unravels the predictions on a moving periodic total (MPT) series. For decimal, you can specify any decimal value from 0.2
through 1.0
as long as the difference between the values of MPTDECAY MIN and MPTDECAY MAX is evenly divided by 0.4
. The default value of decimal is 1.0
.
MPTDECAY MIN decimal specifies the minimum value of the parameter that the forecasting engine uses when it adjusts the decay of estimates of base values that it uses when it unravels the predictions on a moving periodic total (MPT) series. For decimal, you can specify any decimal value from 0.2
through 1.0
as long as the difference between the values of MPTDECAY MIN and MPTDECAY MAX is evenly divided by 0.4
. The default value of decimal is 0.2
.
Specifies the number of trials that the forecasting engine runs to determine the forecast. For integer, you can specify any INTEGER
value from 1
through 3
. The default value of decimal is 3
.
Specifies either the number of periods for a single cycle or the number of periods in each of a set of nested cycles.
You do not have to specify this parameter when you are using a dimension of type DAY, WEEK, MONTH, QUARTER, or YEAR. In this case, the forecasting engine derives the periodicity from the number of time dimension periods that constitute a year (for example, there are 52 WEEK periods in a year).
When you are not using a dimension of type DAY, WEEK, MONTH, QUARTER, or YEAR, the default value for cycle-spec is 1
, which specifies that the data is not grouped at all (that is, each period is logically independent).
Cycles are groupings of time periods that repeat through the time span of the data. For example, daily periods can be grouped into a weekly cycle and weekly periods can be grouped into a yearly cycle. In this case, the cycles are said to be nested, with the yearly cycle more aggregate than the weekly cycle, and the weekly cycle more detailed than the yearly cycle. By specifying cycles at a more detailed level, you allow OLAP to conduct a finer-grained search for factors that affect the data.
To specify a single cycle, set cycle-spec to an INTEGER
from 1
through 25000
. The INTEGER
indicates the number of periods into which the cycle should be divided. For example, the INTEGER 12
specifies that the cycle should be divided into 12 periods.
To specify a series of nested cycles, set cycle-spec to a series of up to six INTEGER
values enclosed in parentheses and separated by commas. Each value in the series is the number of periods in one of the nested cycles. The cycles are ordered from most aggregate to least aggregate. For example, when cycle-spec is (52,7)
, this indicates two cycles in which the most aggregate cycle is divided into 52 periods and each of those periods is divided into seven periods. In this example, the year is divided into 52 weeks, and each of those weeks is divided into seven days.
Specifies the ratio of the size of the window that the forecasting engine uses for smoothing and the total number of historical periods. The forecasting engine uses this value to determine the number of backcast periods. You can specify any decimal value from 1/26
through 1/2
. The default value of decimal is 1/3
.
Indicates whether the forecasting engine should smooth the data for the forecast. The default value is NO
. Specify YES
when you want the forecasting engine to smooth the data.
The data filter that the forecasting engine uses when executing the forecast.
'TRNOSEA' indicates that the forecasting engine will not seasonally adjust the data. (Default)
'TRSEA' indicates that the forecasting engine will transform using a filter that seasonally adjusts the data.
'TRMPT' indicates that the forecasting engine will transform using a moving periodic total (MPT) filter.
For the double exponential smoothing and Holt-Winters forecasting methods, specifies the value of the trend hold parameter that indicates trend reliability for those methods.
TRENDHOLD MAX decimal specifies the maximum value of the trend hold parameter. For decimal, you can specify any decimal value from 0.0
through 1.0
. The default value of decimal is 0.8
.
TRENDHOLD MIN decimal specifies the minimum value of the trend hold parameter. For decimal, you can specify any decimal value from 0.0
through 1. 0
. The default value of decimal is 0.4
.
TRENDHOLD STEP decimal specifies the value of the interval that the forecasting engine uses when it determines the value of the trend hold parameter. For decimal, you can specify any decimal value from 0.1
through 0.2
. The value of decimal must evenly divide the difference between the values of TRENDHOLD MAX and TRENDHOLD MIN. The default value of decimal is 0.2
.
Specifies the number of points that the forecasting engine uses when it determines median values when it performs median smoothing. Median smoothing eliminates extreme variations in the data by replacing each data point in a series by the median value of itself and its neighbors. For integer, you can specify any INTEGER
value from 1
through 13
. The default value of integer is 3
.
For an example of a forecasting program, see Example 14-9, "A Forecasting Program".