Sistema de recomendação de benefícios para famílias inscritas em programas socioassistenciais no Estado do Ceará

Abstract

Currently, social assistance programs in the state of Ceará consult the Unified Registry (CadÚnico) database the Cadastro Único (Single Registry) database as the sole instrument to verify the suitability of a family for enrollment in cash transfer programs. The search for optimization in the provision of public policies for the citizen is a highly relevant theme within the context of a digital government. However, the immensity and concentration of citizen data is something that makes it difficult to identify the right information for the right people, and consequently, the assertiveness of services. To try to solve these problems, the construction of recommendation systems is a widely used strategy. Therefore, this work aimed to create a benefit recommendation system for families enrolled in social assistance programs in the state of Ceará. From the data collected by an action of the Mais Infância Ceará Card Program (CMIC), a machine learning model was built, which used the information from 30,889 families and is able to provide the probability of the family having a certain benefit, which This can be used as a guide for the decision about granting it. Through this work, it was possible to obtain results in six distinct sections: data preprocessing, feature engineering, feature selection, performance evaluation of the models on the training set, performance evaluation of the models on the test set, and evaluation of the recommendation system. It is concluded through this, that there is a great potential in using the data, extracted from the CMIC forms, to optimize the delivery of benefits to families in the programs. Furthermore, the model created is able to recommend a specific benefit, for new records, or records unknown to it. This recommendation, in the current scenario, even serves to compare the existing concessions with the suggested recommendations