EDOBE XDOM PMML Bedienungsanleitung PDF herunterladen (Seite 52)

Factor

The factor data type is much more complicated. A factor column

maintains a list of strings, representing the levels of the factor. When a

string is read from a file or database into a factor data column (by the

file reader proc), each string value is compared to the list of level

strings, and converted into a level number (1 through the number of

levels).

If you know all of the possible factor levels that can be read ahead of

time, you can simply set up a factor column with these levels.

However, if you don't know all of the possible factor levels, there is a

potential problem. The pipeline is designed to be used for problems

with very large amounts of data. If every new factor level that

appears is simply added to the level list, then it is possible that the

system would allocate more and more different level strings, until you

run out of memory. Even if the number of possible levels is relatively

small, there are situations where it is useful to restrict it even further.

Some operations such as tree modeling and crosstabs can take

massive amounts of time or space for variables with many factor

levels.

To control the number of automatically-created factor levels, each

column has two properties,

max.auto.levels and overflow.level.

For a factor data type, the

max.auto.levels property determines the

maximum number of levels that will be automatically created. The

actual number of levels in a factor can be set larger than this, by

explicitly setting the level strings. The default value of

max.auto.levels is 10. A simple way to disable auto-creation of

factors is by setting

max.auto.levels to 0.

The

overflow.level property is used when handling a new factor

level. If adding the new level would cause the number of levels to

exceed the

max.auto.levels property, the overflow.level string is

used instead. For example, if the levels are

"yes" and "no",

max.auto.levels is 2, and overflow.level is "yes", then all other

factor values will map to

"yes". If the overflow level is not one of the

existing levels, it is added, but it is done soon enough so that the total

number of levels will not exceed

max.auto.levels. If

overflow.level is "", the default, then overflow levels are mapped to

NA.

1 2 ... 47 48 49 50 51 52 53 54 55 56 57 ... 97 98

Kommentare zu diesen Handbüchern

Keine Kommentare

EDOBE XDOM PMML Bedienungsanleitung Seite 52

Kommentare zu diesen Handbüchern