Genetic Algorithms for Simultaneous Equation Models

摘要

Traditionally, Simultaneous Equation Models (SEM) have been developedby people with a wealth of experience in the particular problem represented by themodel. To obtain automatically a satisfactory SEM would help the experts in the de-velopment of such models. This paper presents an algorithm to obtain a satisfactorySEM from a set of variables. Because of the extensive number of SEM possible, exhaus-tive search methods are not well suited here. The algorithm combines genetic and arandom search method. The behavior of the algorithm is studied and a parallel versionin shared memory is developed and studied.