Integrating campaign management and data mining

Details Integrating campaign management and data mining Integrating campaign management and data mining

1998-06-15

2001-05-29

Amsbury, Wayne (Department: 2171)

Data processing: database and file management or data structures

Database design

Data structure types

C707S793000

Reexamination Certificate

active

06240411

ABSTRACT:

FIELD OF THE INVENTION
The invention relates to methods and systems for analyzing and selecting records stored in a computer database, and more particularly, to methods and systems for integrating the modeling of new characteristics of records and selection of records from a database.
DISCUSSION OF THE RELATED ART
Computer databases have proliferated. For example, extremely large databases (or “data warehouses”) have been generated for marketing data. While this data may be easy to compile (in some applications), using the information to achieve a goal can be challenging.
A database may be thought of as including one or more tables, with rows of the table corresponding to individual records in the database. For example, in the database
13
of
FIG. 1A
, the first row
19
a
indicates the labels for fields of the overall table
15
. The term “table” refers to any group of associated records, whether stored in actual table format or otherwise. Each of the rows
19
b
-
19
e
is an individual record corresponding to an individual person (in this example). The term “record” includes any associated set of fields (e.g. the fields in row
19
b
of FIG.
1
A). Thus, in row
19
b
, a person has a name “1” (column
17
a
), an age of 65 (column
17
b
), and an income of $60,000 (column
17
c
). While the description of the invention may be provided with reference to records (e.g.
19
b
-
19
e
of
FIG. 1A
) within a table
15
in a database
13
, this is not intended as limiting. The present invention has application for analysis and selection of stored information in a database, no matter what the particular internal representation is. The database may be digital information stored in any digital storage medium, such as conventional random access memory, tape storage, CD-ROM, and others.
The database may be built using a great variety of information for each corresponding component or record of the database. For example, in a database where the records correspond to individuals, the individual's age, address, and income may be readily available information for input to the database. These individual fields, however, may not be meaningful for determining any action in the future. For example, if a business wishes to send out a targeted mailing, the business would like to estimate which of the records in a database of individuals corresponds to individuals who are likely to respond favorably to the targeted mailing. Simple analysis of the available fields in the database (e.g., age, income, and others) may not be sufficient to perform this task.
Accordingly, a number of techniques have been developed for manipulating the known fields (i.e., the characteristics recorded in the database, corresponding to the columns
17
a
-
17
c
, i.e., name, age, and income) to determine a new characteristic (e.g., field) that is more meaningful. Such techniques include those referred to in the art as “data mining.”
FIG. 1B
illustrates one way (of a number of ways) of developing a new field for the database. A database
10
is provided that includes both training data
11
and test data
12
. The training data
11
is a table (including a number of records). The training data is provided to a model builder
14
. The model builder
14
may be software running on general purpose computer. Examples of commercially available packages that may be used for a model builder
14
include: Enterprise Miner (and standard SAS modeling software found in SAS/Base, SAS/STAT, etc.), available from the SAS Institute (“SAS”) of Cary, N.C.; the SPSS program available from SPSS of Chicago, Ill.; Intelligent Miner available from IBM of Armonk, N.Y.; Darwin, available from Thinking Machines of Burlington, Mass.; Modell, available from Unica of Lincoln, Mass.; NeuralWorks Predict, available from NeuralWare, of Pittsburgh, Pa.; and MineSet, available from Silicon Graphics of Mountain View, Calif. The model builder
14
may also be a custom or semi-custom design for implementing a model, such as a hardware implementation of a neural-network.
The model builder
14
constructs a model
16
. The model
16
may be some general method or technique for computing a new value or other parameter based on one or more fields within the record of the training data
11
. The model
16
may, for example, be a statistical analysis or mathematical equation for computing a probability (for example, the probability that a customer would respond favorably to a mailing), a true/false field, or any other numerical, alphanumeric or other result. The model
16
may also produce more than one field. For example, the model might result in calculation of both a determination that a market is going up and a confidence level that the market is going up.
The result of the model or new field may be referred to as a “score.” Thus, the table
33
and the database
30
includes a column
32
entitled “score.” This score may have been determined by a model that was built according to the technique generally illustrated with respect to FIG.
1
.
Once the model builder
14
has arrived at a model
16
, an evaluator
18
may then assess the usefulness of the model
16
. This may be done by examining the results of application of the model
16
to a separate database table
12
that includes test data, stored in the database
10
.
The evaluator
18
may also be a software module implemented on a general purpose computer. Existing software to perform this function is known in the art. SAS and SPSS, described above, are both general statistical tools that can be used to evaluate a model. In addition, many data mining tools (including most of these listed above) also have evaluation functionality built into the software, and may be incorporated as a common part of a software package with the model builder
14
.
Once a model has been constructed and selected for use, as generally described above with reference to
FIG. 1B
or in some other matter, the model may be applied to other databases.
FIG. 2
illustrates an example of application of a model to a database
20
. The model
26
is fed to a model engine
22
. For example, the model
26
may be an executable file that can be applied by the model engine
22
. The model engine
22
takes as an input a database
20
. The database
20
may be a database such as that shown at
15
, but including only columns
17
a
-
17
c.
The model engine
22
may then apply the model to each record in the database
20
to produce a modified database
24
. This modified database
24
would include the results of application of the model
26
to the database
20
. Thus, the modified database
24
could be a table that includes an extra field (or column) that specifies the results of application of the model (or a separate table storing a key and a score, which can be joined with other tables). For example, in table
33
of
FIG. 3
, application of the model
26
could have resulted in a score, which is added to the table
33
in the last column
32
d
. Each record (e.g.
31
b
-
31
e
) receives a corresponding calculated model value (e.g. rows
31
b
-
31
e
in column
32
d
).
FIG. 3
illustrates one method of using a database in order to achieve certain goals. The database
30
includes a table
33
. The table
33
may be as generally described above, i.e., including a number of individual records corresponding to persons (households, businesses, entities or anything else), e.g., rows
31
b
-
31
e
, in a number of fields for each record (corresponding to
32
a
-
32
d
). (While illustrated as a single table
33
, this may actually be stored in multiple tables, joined by a key.) One or more of the fields may correspond to a characteristic computed according to one of the above models generated through data mining or other technique, e.g. column
32
d
having a score.
The table
33
may be provided to a campaign manager
34
. The purpose of campaign management is to select and categorize the records of the database (e.g., a corresponding row, such as
31
b
,
31
c
,
31
d
or
31
e
) for a variety of actions (or create a “segment” or segm

Affiliated with

Also associated with

Rating

Integrating campaign management and data mining does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Integrating campaign management and data mining, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Integrating campaign management and data mining will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2523164