Modelo de otimização de políticas de crédito para instituições brasileiras de microfinanças

Abstract

This thesis built a computational model for decision making support in credit loan for Microfinance Institutions (MFI). One developes a multi-objective optimization tool to maximize simultaneously both levels of efficiency (measured as the selection of candidates for creditloan with lower default probability) as well as efficacy (a proxy to portfolio credit volume) for credit loan candidates selection policies. The objective is to achieve self-sustaining without the necessity of external donations. The main methodological contribution is an application of a Monte Carlo technique for permutation tests to generate configurations under the nullhypothesis in multi-objective problems, instead of employing the most frequently technical applied in the computational economy and financial areas, such as classical econometric tools and DEA. It was built the significance vector concept in order to infer the robustness of theresults found within the Pareto optimal front. The ability to select, through quantitative criteria, the most adequate policy amongst the Pareto front solutions, without resorting to arbitrary weights to balance the relative importance of the mono-objective functions of efficiency and efficacy, makes this method a novelty in credit policies evaluation area.