摘要：Currently, most processes have a volume o f histor- ical information that makes its ma nua l processing difficult. Data mining, one o f the most significant sta ges in the Knowledge Discovery in Databases (KDD) process, has a set of techniques capable o f modeling and summarizing these historic al data, making it easier to unde rsta nd them and helping the decision making proc ess in future situations. This article presents a new data mining adaptive technique ca lle d SOM+PSO that can build, from the availa ble information, a reduce d set o f simple classific ation rules from which the most significant rela tions b e- tween the features recorded can be derived. These rules operate both on numeric and nominal attributes, and they are built by combining a variation of a population metaheuristic and a competitive ne ural network. The method proposed was compare d with the PART method and measure d over 19 databases (mostly from the UCI repository), and satisfactory re sults were obtained.
关键词：Classification Rules; Data Mining; Adaptiv e ; Strategies; Particle Swarm Op timization; Self-Organizing ; Maps.
