Community Research and Development Information Service - CORDIS


In this work, it is shown how to extract knowledge, as a rule set, from data by means of a genetic algorithm using a procedure which deals well with most problems associated with real datasets. As their name suggests, genetic algorithms replicate the main natural selection mechanisms of selection, reproduction and mutation. This characteristic allows them to be a powerful optimization tool. In particular, it can largely avoid being trapped into local extrema when searching a global one in complex multivariable spaces. The report is structured as follows. In a first part, genetic algorithms are described in a more detailed way. One particular way of applying them to knowledge extraction is then shown with special emphasis on real world datasets. Results of this approach are given in the fourth part with two examples, namely, credit risk assessment and prediction of transport user behaviour.

Additional information

Authors: CANARELLI P, JRC Sevilla (ES)
Bibliographic Reference: EUR 16446 EN (1996) 17pp., FS
Availability: Available from the Public Relations and Publications Unit, JRC Ispra, I-21020 Ispra (IT), Fax: +39-332-785818
Record Number: 199611441 / Last updated on: 1996-12-16
Original language: en
Available languages: en