EDOBE XDOM PMML Bedienerhandbuch PDF herunterladen (Seite 10)

FPGROWTH Algorithms Renamed and Modified

supported. The table below shows sample calls for both the older and new algorithm

names.

Prior Release (old way) Release 2.0 (new way)

CALL nza..PREPARE_FPGROWTH(

'intable=nza..quant_sales, outtable=dset, tid=tid,

item=idart');

CALL nza..PREPARE_ARULE(

'intable=nza..quant_sales, outtable=dset, tid=tid,

item=idart');

CALL nza..FPGROWTH('intable=nza..retail,

pfx=results, support=1 ');

CALL nza..ARULE('intable=nza..retail,

model=assoc, supporttype=percent, support=5,

lvl=0, maxsetsize=5, confidence=0.5');

Detailed information about these algorithms can be found in the IBM SPSS In-Database

Analytics Developer's Guide and the IBM SPSS In-Database Analytics Reference Guide.

Modified Default Parameter Settings for DECTREE and

REGTREE

Default parameter settings for decision and regression tree growing are modified in this re-

lease to yield smaller trees. The previous defaults resulted in creating large trees, maximal-

ly fitted to the training data (good candidates for pruning). Growing, pruning, and using

such large trees for prediction is computationally expensive and often unnecessary. By us-

ing modified parameter settings, you can obtain substantial computational savings. The al-

gorithms are usually run with their default settings, so those defaults are changed to pro-

duce smaller trees. To force large tree growing, those users with an understanding of the

performance impact can change the parameter settings.

Decision Trees

The following default parameter settings are changed:

Parameter New Value Old Value Description

minsplit 50 2 The minimum number of instances in a node required for a split. If

the number of instances in a node is less than minsplit, no further

split is applied and the node becomes a leaf.

maxdepth 10 62 The maximum decision tree depth. If a node's level in the tree

equals maxdepth (with the level of the root node equal 1, the level

of its descendants equal 2, etc.), no further split is applied and the

node becomes a leaf.

minimprove 0.02 0.01 The minimum improvement of the evaluation function used for

split selection (class impurity measure) required for a split. If the

decrease of class impurity (calculated using the function specified

via the eval argument, defaulting to the entropy) is less than min-

improve, no further split is applied and the node becomes a leaf.

Regression Trees

The following default parameter setting is changed:

1 2 ... 5 6 7 8 9 10 11 12 13 14 15 ... 27 28

Kommentare zu diesen Handbüchern

Keine Kommentare

EDOBE XDOM PMML Bedienerhandbuch Seite 10

Kommentare zu diesen Handbüchern

Verwandte Produkte und Handbücher für Nein EDOBE XDOM PMML