Memory allocated megs: 2
NOTE: The PROCEDURE SEQUENCE used 0:00:33.42 real 0:00:16.17 cpu.
Copyright 2000 by SAS Institute Inc., Cary, NC, USA. All rights reserved.
The NITEMS= option specifies the maximum number of events for which
rules, or chains, are generated.
The SAME= option specifies the lower time-limit between the occurrence
of two events that you want to associate with each other (default = 0).
visit time / same=2;
The SORT procedure sorts the observations in descending order by the
values of support.
proc sort data=s4out;
by descending support;
The PRINT procedure lists the first 10 observations in the sorted sequence
proc print data=s4out(obs=10);
var count support conf rule;
title 'Partial Listing of the 4-Item Sequences';
title2 'Lower Timing Limit Set to 2';
Agrawal, R., Imielinski, T., and Swami, A. (1993), "Mining Association Rules between Sets of
Items in Large Databases", Proceedings, ACM SIGMOID Conference on Management of Data,
207-216, Washington, D. C.
Berry, M. J. A. and Linoff, G. (1997), Data Mining Techniques for Marketing, Sales, and
Customer Support, New York: John Wiley and Sons, Inc.
PROC SPLIT Statement
Example 1: Creating a Decision Tree with a Categorical Target (Rings Data)
Example 2: Creating a Decision Tree with an Interval Target (Baseball Data)
An empirical decision tree represents a segmentation of the data created by applying a series of simple
rules. Each rule assigns an observation to a segment based on the value of one input. One rule is applied
after another, resulting in a hierarchy of segments within segments. The hierarchy is called a tree, and
each segment is called a node. The original segment contains the entire data set and is called the root
node of the tree. A node with all its successors form a branch of the node that created it. The final nodes
are called leaves. For each leaf, a decision is made and applied to all observations in the leaf. The type of
decision depends on the context. In predictive modeling, the decision is simply the predicted value.
Besides modeling, decision trees can also select inputs or create dummy variables representing
interaction effects for use in a subsequent model, such as regression.
PROC SPLIT creates decision trees to either:
classify observations based on values of nominal or binary targets,
predict outcomes for interval targets, or
predict the appropriate decision when decision alternatives are specified.
PROC SPLIT can save the tree information in a SAS data set, which can be read again into the
PROC SPLIT can apply the tree to new data and create an output data set containing the predictions, or
the dummy variables for use in subsequent modeling. Alternatively, PROC SPLIT can generate DATA
step code for the same purpose.
Tree construction options include the popular features of CHAID (Chi-squared automatic interaction
detection) and those described in Classification and Regression Trees(Breiman, et al. 1984).
For example, using chi-square or F-test p-values as a splitting criterion, tree construction may stop when
the adjusted p-value is less significant than a specified threshold level, as in CHAID.
When a tree is created for any splitting criterion, the best sub-tree for each possible number of leaves is
automatically found. The sub-tree that works best on validation data may be selected automatically, as in
the Classification and Regression Trees method. The notion of "best" is implemented using an
assessment function equal to a profit matrix (or function) of target values.
Decision tree models are often easier to interpret than other models because the leaves are described
using simple rules. Another advantage of decision trees is in the treatment of missing data. The search
for a splitting rule uses the missing values of an input. Surrogate rules are available as backup when
missing data prohibit the application of a splitting rule.