Discovering new orders of the chemical elements through genetic algorithms

Abstract

The design of new materials is a major issue in many domains (electronics, environment and so on). A large number of databases have been developed in order to help scientists to design new materials. Databases of experimental results can be used to learn prediction models of each property. Data mining methods, can be applied on such databases to discover empirical rules and predict properties. In this paper we propose a method for discovering new orders of the chemical elements. This reorganization of the chemical elements can be used to improved prediction accuracy of classification methods and to enhance similarities between elements. A genetic algorithm is used to find a satisfying solution according to several evaluation criteria through a Pareto-based multi-objective approach. We carried out several experiments of prediction of compound formation (ternary chalcopyrite compounds A BX2, where X is either S,Se or Te).The first results showed that distance-based evaluation seems promising, as it has been possible to discover groups of similar elements regarding the task. © 2008 Springer-Verlag Berlin Heidelberg.