%0 Conference Paper %1 pappa:2004:ecml %A Pappa, Gisele L. %A Freitas, Alex A. %B ECML/PKDD 2004 Proceedings of the Workshop W8 on Advances in Inductive Learning %C Pisa, Italy %D 2004 %E Furnkranz, Johannes %K classification data genetic induction, mining, programming, rule %P 93--108 %T Towards a genetic programming algorithm for automatically evolving rule induction algorithms %U http://www.cs.kent.ac.uk/pubs/2004/2023 %X Rule induction is one of the techniques most used to extract knowledge from data, since the representation of knowledge as if/then rules is very intuitive and easily understandable by problem-domain experts. Existing rule induction algorithms have been manually designed. In this era of increasing automation, Genetic Programming (GP) represents a powerful tool for automatically evolving computer programs. This work proposes a genetic programming algorithm for automatically evolving rule induction algorithms. Hence, the evolved rule induction algorithm will, to a large extent, be free from the human biases that are implicitly incorporated in current manually-designed algorithms (such as the typical use of a greedy search method). This is a very ambitious, adventurous goal, which, if successful, will pave the way for a new generation of more robust, considerably less greedy rule induction algorithms. In particular, an automatically evolved rule induction algorithm can be designed to cope with attribute interaction better than current greedy rule induction algorithms, which will tend to lead to an improved performance in complex data sets.

@inproceedings{pappa:2004:ecml, abstract = {Rule induction is one of the techniques most used to extract knowledge from data, since the representation of knowledge as if/then rules is very intuitive and easily understandable by problem-domain experts. Existing rule induction algorithms have been manually designed. In this era of increasing automation, Genetic Programming (GP) represents a powerful tool for automatically evolving computer programs. This work proposes a genetic programming algorithm for automatically evolving rule induction algorithms. Hence, the evolved rule induction algorithm will, to a large extent, be free from the human biases that are implicitly incorporated in current manually-designed algorithms (such as the typical use of a greedy search method). This is a very ambitious, adventurous goal, which, if successful, will pave the way for a new generation of more robust, considerably less greedy rule induction algorithms. In particular, an automatically evolved rule induction algorithm can be designed to cope with attribute interaction better than current greedy rule induction algorithms, which will tend to lead to an improved performance in complex data sets.}, added-at = {2008-06-19T17:35:00.000+0200}, address = {Pisa, Italy}, author = {Pappa, Gisele L. and Freitas, Alex A.}, biburl = {https://www.bibsonomy.org/bibtex/21cdab2e5b651ece1d90cc0abcf28fb7d/brazovayeye}, booktitle = {ECML/PKDD 2004 Proceedings of the Workshop W8 on Advances in Inductive Learning}, editor = {Furnkranz, Johannes}, interhash = {5d50aa295325bedfa7a06a01888c1d07}, intrahash = {1cdab2e5b651ece1d90cc0abcf28fb7d}, keywords = {classification data genetic induction, mining, programming, rule}, month = {20-24 September}, notes = {http://www.cs.kent.ac.uk/pubs/2004/2023/index.html http://www.ke.informatik.tu-darmstadt.de/events/ECML-PKDD-04-WS/}, pages = {93--108}, publication_type = {inproceedings}, refereed = {yes}, size = {16 pages}, submission_id = {4571_1102851561}, timestamp = {2008-06-19T17:49:13.000+0200}, title = {Towards a genetic programming algorithm for automatically evolving rule induction algorithms}, url = {http://www.cs.kent.ac.uk/pubs/2004/2023}, year = 2004 }