This paper presents data mining (DM) solutions based on evolutionary methods. The framework emphasizes the suitability of genetic algorithms and genetic programming in data mining context. We first describe the concepts and their closed links with machine learning (ML) and statistics. Two main data mining tasks are considered: the classification and association analysis. While classification has been intensively studied in ML, association analysis is typically related to DM; both may be achieved efficiently with genetic-based methods. A clear distinction between these two data mining functionalities, which result in syntactically comparable patterns, is established. The genetic-based techniques used in DM context are presented. We show how individuals, genetic operators and fitness functions are mapped in order to address the specific database issues. Suitable characteristics to database analysis are pointed out and research challenges presented.
展开▼
