Um algoritmo evolucionário para encontrar sequências de transformações com ordenação por classe

Abstract

Evolutionary Algorithms are metaheuristics widely used to solve combinatorial problems and optimizational. A combinatorial problem, on the context of Computer Science, is the choose for a transformations sequence that is used by the compiler during the target code generation. This work aims to propose and evaluate an evolutionary algorithm capable of find effective transformations sequences, that is potentialized by a classification, which respect an application order of transformation classes. The results indicates that the proposed technique is efficient to find good transformations sequences, besides to be a good option to databases generation to machine learning systems.