Aplicação de Inteligência Artificial na gestão do patrimônio imobiliário público do Instituto Nacional do Seguro Social

Resumo

This study presents possible financial impacts resulting from the use of Artificial Intelligence (AI) in the management of federal public real estate assets, by comparing forecasts of rental appraisal values obtained by Artificial Neural Network (ANN) Algorithms with the results of Multiple Linear Regression Models (MLR), in cases of third-party properties leased to the National Institute of Social Security (NISS). The objective of this Technological Report was to build a Protocol and/or Methodology to assist Public Management decisions regarding the values of negotiated properties, by comparing the performance of the traditional statistical inference theory, used by federal real estate valuation engineering, to mathematical algorithms similar to biological neural structures. Considering the possibility of reducing, or improving the management, of the current amount of approximately R$ 48 million spent annually on rent by the NISS, and the High Risk scenario in which the Public Administration of Federal Real Estate Assets finds itself, the results indicate that the RNA Algorithms: Levenberg-Marquardt, Batch Backpropagation, Recursive Backpropagation, RPE – Forgetting Factor, RPE – Constant Trace and RPE – EFRA, serve as a tool in decision-making regarding properties negotiated by the public administration, since, the type of RNA Algorithm, and considerations regarding the packaging (Bagging) of the sample data, provide lower results of up to -13.6% for rental values of third-party properties for use by the Administration, when compared to RLM models, suggesting financial impacts in the use of AI in the management of federal real estate assets, in addition to scientifically substantiating the justifications for the use of the Field of Arbitration on market rental values, enabling the reduction of costs/expenses, in cases where the Administration rents (or buys) properties from third parties, and the increase in revenue/collection, in cases where the Administration rents (or sells) its own properties to third parties.

Ficha do documento

Tipo

Dissertação

Ano

2024

Instituição

Fundação Getulio Vargas

Fonte

Repositório da FGV

Idioma

Português

Acesso

Acesso aberto

Identificador

oai:repositorio.fgv.br:10438/36482

Temas

Gestão PúblicaAdministração PúblicaTecnologia

Palavras-chave

Redes Neurais ArtificiaisRegressão Linear MúltiplaAdministração PúblicaPatrimônio ImobiliárioInteligência ArtificialPropriedade públicaRedes neurais (Computação)Instituto Nacional do Seguro Social (Brasil)