Proposição de modelos de previsão de risco de crédito para pequenas e médias empresas por meio da regressão logística

• Ano de defesa: 2018
• Autor(a) principal: Führ, Flávio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Pato Branco; Brasil; Programa de Pós-Graduação em Engenharia de Produção e Sistemas; UTFPR
• Palavras-chave em Português: Análise de crédito; Sistema de avaliação de risco de crédito (Finanças); Pequenas e médias empresas; Credit analysis; Credit scoring systems; Small business; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO; Engenharia de Produção
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/3582

The present work seeks to contribute to the financial sector of credit promotion for small and medium enterprises (SMEs). For this purpose, it proposes the elaboration of credit risk forecasting models for SMEs through Logistic Regression (RL). Using data from credit registry and credit history, variables with significant significance were extracted for the definition of probability of occurrence of default. Data collection and generation of information was done through the exploratory research, with experimental procedures, having as the field of exploitation a credit cooperative database. Methodologically, four classes of companies were created: Individual Microentrepreneur (MEI), Microenterprise (SM), Small Business (PE) and Medium Enterprise (MédE). The general database was redistributed according to the billing ranges. the 4 new databases. To improve the models and reduce the differences, within the database, of each class of companies, the process of discretization and the creation of dummy or artificial variables was used. As a result of the application of the statistical technique in the database, in the 4 classes of companies: MEI, ME, PE, MédE and in the General Data (DG), a confirmation of the relevance of the RL in the elaboration of the models was obtained. The accuracy of the models presented expressive percentages for the database with non-accounting and non-auditable variables, reaching satisfactory percentages. For MEI, the percentage of accuracy was 83%, using 2 variables in the composition of the model. As for ME, it presented an accuracy of 84.9%, using 5 variables in the model composition. For PE the accuracy reached 88.5%, however including only 1 variable in the model. For MEc the accuracy was 83%, presenting 3 variables in the model and for the DGs, the accuracy was of 85%, presenting 5 variables in the composition of the model. It was still possible to observe which variables have greater relevance within the database. The models developed are tools that can contribute to the credit analyst in the identification of possible good payer or defaulters for financial institutions that have SMEs in their portfolio.