MODELOS DE PREDICCIÓN DE DESERCIÓN DE CLIENTES PARA UNA ADMINISTRADORA DE FONDOS ECUATORIANA

Abstract

The existence of a company is justified by its customers, who are active as the most important assets. Faced with more competitive markets and where the needs of customers are increasingly demanding, companies seek efficiency in the use and analysis of data. Losing customers is more expensive than attracting new customers. The study on customer behavior, specifically attrition, has become a prevailing need within the business environment. In the presentation of research, data mining techniques are used to build models of customer attrition prediction, which can be applied within the financial disintermediation market. The statistical models used are: Decision Trees, Random Forests and Logistic Regression, these are evaluated in terms of accuracy by the area below the receiver operating characteristics curve (ROC). The evaluation of the results, the evaluation that the random forest has a better performance than the other models applied in the study.