Exploitation of modern heuristic techniques within a commercial data mining environment.
The development of information technology allows organisations
to gather and store ever increasing quantities of data. This data,
although not often collected specifically for such a purpose, may be
processed to extract knowledge which is interesting, novel and useful.
Such processing, known as data mining, demands algorithms which
can efficiently 'mine' through the large volumes of data and extract
patterns of interest. Modern heuristic techniques are a class of optimisation
algorithms, which solve problems by searching through the
space containing all possible solutions. They have been applied to
a wide variety of such problems with great success, which suggests
that they may also prove useful for data mining. Conducting a search
through the space of all patterns within a database using such techniques
is likely to yield useful information.
Within this thesis, it is demonstrated that modern heuristic techniques
may be successfully applied to a wide range of data mining
problems. The results presented highlight the suitability of such algorithms
for the demands of the commercial environment; as a consequence
of this, much of the work undertaken has become incorporated
within real business processes, bringing considerable savings.
A variety of algorithmic enhancements are also investigated, yielding
important results for both the data mining and heuristics fields.