Constructing X-of-n Attributes With A Genetic Algorithm
Otavio Larsen, Alex Alves Freitas, Júlio César Nievola · 2002
Abstract: We propose a new Genetic Algorithm (GA) for constructive induction. The goal of the GA is to construct new X-of-N attributes out of the original attributes of a given data set, in order to improve the effectiveness of a data mining algorithm to be applied to that data set. The GA follows the preprocessing approach for constructive induction, so that the new attributes constructed by the GA can be given (together with the original attributes) to any data mining algorithm. In this paper we use as a data mining algorithm C4.5, which is a well-known decision-tree induction algorithm. In order to evaluate the effectiveness of the new attributes constructed by the GA, we compare the performance of C4.5 with and without the new attributes constructed by the GA across several data sets.