Maestría en Matemática Aplicada

Permanent URI for this collectionhttp://repositorio.uta.edu.ec/handle/123456789/32203

Browse

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Modelo matemático para medir la probabilidad de la incurrencia en mora de créditos utilizando regresión lineal en una institución financiera de la ciudad de Ambato
    (Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Maestría en Matemática Aplicada, 2022) Marcalla Pilamunga, Luis Alberto; Bustamante Romero, Edgar Johni
    This paper deals with the analysis of credit concessions in financial institutions in the city of Ambato, in order to estimate the probability of return on investment, reducing risk and granting agreements between the borrower and the lender. The intrinsic characteristics of the research, of the predictor variables and the independent variable that were used resulted in the application of a generalized linear model with the binomial logit option. The predictor variables that for the present study are all categorical, became fictitious or dummy variables when incorporated into the model, and the dependent variable, which can take only two possible values, 0 or 1, the same one that estimates the probability of default; grant to the present investigation a different approach that considers the socioeconomic, social, territorial and even gender environment to predict if a loan holder will diligently comply with the payments in the agreed installments. The institution that lent its collaboration for the design of the model was the "Cooperative of savings and credit Chibuleo Ltda.", located in the center of the city of Ambato, and in force in the financial market for the last 16 years. The cooperative belonging to the popular and solidarity economy sector has been promoting the local economy through the products and services it offers. The methodology applied in the cooperative to check if a member is a good payer is carried out with the help of different bureaus and with the thoroughness of a credit analyst, who decides whether or not to grant a loan; This methodology is useful, however, the history of loans granted with the same characteristics is not considered. The model made fills that gap and gives us a better overview of the possible scenarios in the future. The data was extracted from the cooperative's database, a relational database in SQL Server 2014, which was loaded into statistical software for processing and analysis. R is used, a programming language focused on a statistical analysis of large volumes of data, in addition to a graphic environment RStudio version 2020.02.1. As the variables studied were categorical, and not continuous, a version of linear regression, logit regression, was used to calculate the probability of relapse into default: 1 to indicate the occurrence of default on a loan bonus and 0 to predict loans without news in payments. For the development of the model, the maximum likelihood method is used with the help of R and the variables were tested to determine their significance with a confidence level of 95%.