WinTree is a system that allows people to specify complex selection
and
segmentation
rules that are to be applied to a large database. The rules
are developed using a graphical editor. Once developed, the rules
are passed to a
mining engine that applies the
rules to the database. The process can produce statistical data,
data for analysis and data for other purposes such as mailing and
personalization.
Multiple sets of
rules can be run at the same time with only one pass of
the database. In addition, WinTree has tightly integrated advanced
data analysis tools. The tools can be used to improve the accuracy
of the selection rules.
WinTree has four main features that address the problems of complex selection:
- A graphical and syntactic language to express the rules that
define sets of names and the relationships between those sets.
The language also allows the user to specify how the names in
a set are to be treated by letting the user indicate what actions
are to be taken on the names.
- A dictionary system that allows rules to be represented by user chosen names.
These names constitute a set of corporate business definitions that can
be shared by many people. They help to provide a uniform and
consistent set of selection rules. Named rules may be combined in
various ways to create more complex rules. For example, in a
marketing context, one could have a rule called SportsAffinity and another
rule called RecentPromotion. One could make a new rule by combining
these two rules in the following way:
SportsAffinity AND NOT RecentPromotion
You may also give this new rule a name so that it may easily
be reused and shared. In this way complex rules may be built
up from basic components. The logic behind the two rules may
be quite complex but, as you can see, they are easy to use.
- A view of data that can readily represent almost any collection
of data with little or no manipulation. We call this view a Micro-database.
It provides a view of a customer's data as a collection of tables.
The data may come from various sources. For example, customer
personal information may come from one file, order/payment history
from another and promotion history from yet another file. In addition,
demographic information could be added. The information is merged
together in memory at selection time to form a composite view
of a customer.
- Built in features that are designed to support the use of analytical
analysis to enhance the selection process. Examples of such features
are the RECODE function and the SAMPLE action. The RECODE function
allows variables to be categorized and also provides a means to
define cross tab variables. The SAMPLE action provides a means
of collecting data for analysis in such a way that the results
of analysis can be easily fed back into the selection system.
|
