Remittance flows estimation hybridizing a multilevel fuzzy system and a simulation model

Abstract

This paper estimates the remittances recorded in Ecuador throughout the time span 2000-2010, when they were the second source of foreign currency of the Andean country. The relationships between covariates, which have effects on these financial flows, are modeled by fuzzy dependence relationships (DR) instead of algebra-based relationships due to the absence or bad quality of data available. A procedure to male fuzzy rules explicit and to evaluate them automatically was designed and tested in a multilevel fuzzy inference engine, which is integrated in a Monte-Carlo simulation model. Firstly, the primary covariates (inputs in a DR) are determined by the simulation model according to expert-based selection of its statistical distributions. Thus, the inference engine evaluated DR outputs once the input values were known, following a hierarchical structure. The results show that this methodology allows us to include expert knowledge in a simulation model.