What Is Data Mining and How Does It Work? - Discrimination and Privacy in the Information Society - page 39

coupling is referred to as integration . The integration of databases leads to a new,
larger database. 26
There are two basic forms of database coupling. First, several records may be
integrated. For instance, suppose that a particular type of cancer is rare, and there
are many hospitals with only a few patients suffering from this disease. Since
every hospital database contains little data, epidemiological research is difficult.
Combining the records of these patients may allow or enhance such research.
Second, attributes may be integrated. In this case, not the number of patients,
but the number of attributes per patient increases. Extending the previous
example, this may become possible when the medical insurance database is
coupled with the medical database of a hospital. Attributes concerning insurance
and medical attributes are now integrated. This can be done, of course, only when
the databases contain information on the same individuals.
The coupling of databases requires that the databases have the same identifier
system, something that is not always the case. 27 However, combinations of
integrating records and integrating attributes are possible as well. In this way,
integration may be full, i.e., all the data are in the new database, or integration
may be partial, i.e., only parts of the data are in the new database. Partial
integration may be used to find missing data. 28

coupling

A B

Two separate databases The coupled database

Fig. 2.5 Separate databases result in far fewer (dual) relationships (represented by lines)
than do coupled databases: two separate databases of four data items (represented by dots)
result in twelve (two times six) relationships (A), whereas the coupled database with a total
of eight data items results in 28 relationships (B).

The different methods of database coupling are illustrated in Figure 2.6. The
blocks in this figure are to be interpreted as relational database matrix structures,
where the rows represent the records and the columns represent the attributes. The
coloured sections are filled with data; the blank sections are empty.
It should be mentioned that in coupling, a distinction is usually drawn between
computer matching, verification, 29 and computer profiling. 30 These forms of
26 Note that the integration may be temporary, since it is possible to retain copies of the
separate databases and to destroy the integrated database after use.
27 National Research Council (1997), p. 118.
28 Running a check of the data against another database is called verification.
29 Verification is sometimes referred to as computer-assisted front-end verification .
30 OTA Report (1986).