A survey and taxonomy of performance improvement of canonical genetic programming
Knowledge and Information Systems
| Hi-index | 0.01 |
Knowledge discovery is the most desirable endproduct of an enterprise information system. Researchesfrom different areas recognize that a new generation ofintelligent tools for automated Data Mining is needed. Inthis sense, this paper explores the Grammatically BasedGenetic Programming (GGP) approach for miningrelational databases, specifically for the classificationtask. Genetic Programming is a powerful inductiontechnique that can be applied to solve a wide variety ofproblems and, in particular, to induce classifiers.Furthermore, knowledge representation using grammarsmakes it possible to represent restrictions, complexstructured objects and relations among objects or theircomponents. A framework named GPSQL Miner wasdeveloped using this approach. It also exploits SQLfeatures to deal with Database Management System(DBMS) that permits fast access to the data. GPSQLMiner automatically generates the input grammar fromsome users' information, like goal and tables, stored inthe DBMS. This grammar is used as bias for the evolutionprocess. In order to validate this approach, the paperpresents results of experiments performed on manydatabases. These experiments show that the proposedapproach is robust, powerful, flexible and able ofattaining good performance.