Construction of a Probability Scoring model for the company SEGUMAR S.A.

Abstract

The objective of this work is the construction of a probability credit scoring model in order to minimize the risk of default on payment of the customer portfolio, for which dependent (customer good and bad) and independent (characteristic of customer) variables were used to provide a correct analysis to determine whether or not the company grants a loan. The descriptive methodology and quantitative and qualitative approaches were applied taking as primary sources the data of the customer portfolio of the company SEGUMAR S.A. The database consists of the information of 100 people applying for a loan and is included in the measurement of 7 variables for each person. Each applicant is classified into one of two possible categories, "good customer" (70 cases) or "bad customer" (30 cases). A credit scoring rule was developed to determine whether a new applicant is a "Good" or "Bad" customer, based on the values of one or more explanatory variables resulting from the final model. This study evaluated the characteristics that customers have at the time of requesting a loan and that according to the characteristics of each customer it is possible to make predictions, classify them as a good customer or a bad customer. In the results obtained from the Logit model it can be concluded that the selected variables that were applied in the model gave us a 76% success rate that allows us to classify each of our customers as a good customer or bad customer in our model.