In science, one can learn the most by studying what seems the least.
Marvin Minsky, The Society of Mind
This is a thesis written for obtaining a Masters degree from the department of Computer and Information Science at the Norwegian University of Science and Technology. I spent four months writing the thesis at Université Paris IX Dauphine, where I was able to come through the Esprit project NOEMIE.
Many people have contributed, inspired the work, and made the writing enjoyable. It started with a project work, An Evaluation of Data Mining Tools and Methods that Endre Lidal and I wrote the spring semester of 1997. We were both interested in Data Mining after following courses in Logic Programming given by professor Jan Komorowski at NTNU. We chose to write a project for NOEMIE to get experience with large projects, and so that our work would lead to something useful. It also made it possible to continue on a thesis later and go abroad.
In the summer of 1997 I was also able to do more practical work on NOEMIE with Ole Martin Winnem at Sintef, and also during the summer, the Principles of Data Mining and Knowledge Discovery conference was arranged in Trondheim.
When I arrived in Paris, I was able to attend a conference on Symbolic Data Analysis and its applications, and later a course on Symbolic Data Analysis given by professor Edwin Diday. Here I was able to give a small lecture on Bayesian networks, which provoked many interesting discussions.
While working on the thesis I wrote an article on Data Mining for the science column in the Norwegian dayli Klassekampen. This lead to a debate with people from other disciplines on ethical consequences and also to a better understanding of the methods.
I have benefited greatly from people who have kindly answered various e-mail queries. I would like to thank Lothar Gierl (University of Rostock), David Aha (Navy AI Labs), Klaus-Dieter Althof (Fraunhofer Institute), Alex Goodall (AI Intelligence), Hideo Shimazu (NEC), Marco Ramoni (Knowledge Media Institute, Open University), Will Taylor (NASA Ames Research Center), Eelco den Heijer (Syllogic), Sunil Vadera (University of Salford), Aubrey Hill, Gregory Piatetsky-Shapiro for allowing me to post a message on the kdnuggets mailing list, and Jack Breese and David Heckerman at Microsoft Research.
At Sintef, I would like to thank Helge Langseth for long e-mail discussions, Helge Sandtorv for help with the OREDA database, and Kjell Corneliussen for evaluating the results obtained from a user perspective.
Many thanks to Per Christian Rangaard at Maison de Norvege for discussions on safety and reliability, and PhD student Frank Olsen for discussions on Case-Based Reasoning and Data Mining in general. Eric Auriol at Acknosoft gave a course in KATE which was very helpful. The PhD students Fredrick Vautrain and Geraldine Polaillon at Dauphine provided interesting discussions. Thanks are also due to fellow students Knut Jørgen Vatnestrøm, Preben Randhol, Tor-Martin Tveit and Jon Stolpnessæter for proofreading.
Finally, I would especially like to thank Moufida Massrali for organizing
my stay at Dauphine. Otherwise, I would probably had lots of extra encounters
with French bureaucracy. Also I would like to thank my supervisors Edwin
Diday and Agnar Aamodt for inspiration and help.
Trondheim, 24.02.98,
Torgeir Dingsøyr