B. de la Iglesia, J. C. W. Debuse, V. J. Rayward-Smth
In this paper we describe our experiences of using simulated annealing and genetic algorithms to perform data mining for a large financial service sector company. We first explore the requirements that data mining systems must meet to be useful in most real commercial environments. We then look at some of the available data mining techniques, including our own heuristic techniques, and how they perform with respect to those requirements. The results of applying the techniques to two commercial databases are also shown and analyzed.