Parallelism on Hybrid Metaheuristics for Vector Autoregression Models

Abstract

Vector Autoregression Models are multi-equation models that linearly describe the simultaneous interactions and behavior among a group of variables, using only their own past. They have been traditionally used in finance and econometrics, but, with the arrival of Big Data, huge amounts of data are being collected in numerous fields and their use for other fields is being considered. Tools are available for these models, but the huge amount of data makes it necessary to exploit High¬Performance Computing for the acceleration of methods to obtain the models. This paper considers a matrix formulation to represent time dependencies, and the solution of the optimization problem generated is approached through hybrid metaheuristics. The parameterized, parallel implementation of the metaheuristics and the matrix formulation ease the exploitation of multilevel shared-memory parallelism.