Database Reference
In-Depth Information
4 Discussions
Uncertainty and Difficulty
Uncertainty is the most noticeable feature of data mining. Given the same data set and
the same objectives, there can be various ways of preparing and mining the data.
Because of the uncertainty, trials of alternatives are unavoidable. The exercise of
sound judgement is therefore essential. The sense of sound judgement depends on
levels of understanding of knowledge, analytic skills and experience. This means that
data mining projects can be seen as time consuming and difficult. However, the diffi-
culty should not come as a surprise, nor should it suggest that the project is unsuitable
for undergraduate students. Data mining should be taught to final year undergraduate
students upon whom a greater degree of academic maturity in terms of analytic skills,
logical reasoning and soundness of judgement is expected. With advices and direc-
tions from the tutor, the difficulty can be overcome.
There is a genuine difficulty, i.e. the absence of domain experts from the data min-
ing lifecycle.
Domain experts
are those who know the application domain well and
can judge which patterns are potentially useful and which are not. The experts are in
fact present throughout the data mining process in most, if not all, real-life data min-
ing projects [4]. It is not realistic to expect the tutor to play such a role for various
domains of application for all kinds of data sets. This particular difficulty could be
avoided by the tutor intervening in the data set selection and only allowing students to
select a data set where the tutor is familiar with the application domain. By doing so,
however, the motivation of the students may be affected.
Usefulness of Case Studies
Because of the uncertainty and difficulties, it is useful and beneficial to study cases of
reported data mining projects. These case studies bring the data mining process alive
and provide students an opportunity to observe how data mining is done before they
try it themselves. The value of case studies cannot be underestimated. Good case
studies were rare and difficult to get. Indeed, industry and commerce often consider
data mining as a closely guarded secret. However, the situation appears changing. It is
now possible to find successful commercial data mining projects ([4], [5]). It is rec-
ommended that a successful case be presented in teaching sessions so that every as-
pect of the case can be discussed thoroughly under the tutor's supervision.
The author has used an assignment as a way of getting good case studies: students
are asked to search for a good case in a designated application area from the published
sources. They are then required to study the case, present it and comment on it. This
work has been proved very useful for students, and can be taken as a part of assess-
ment. Such an assignment may also be considered as a part of the coursework element
for modules that only briefly cover the topic of data mining.
Levels of Expectation and Project Scope
The level of expectation must reflect the aim of the project, i.e. to provide students
with an opportunity to
experience
data mining. It is necessary to emphasise again that
the success or failure of a project should not be judged on applicability of the result
patterns, but on the tasks and actions taken and their justifications.
