Estudo da base de dados abertos E-Saúde da prefeitura de Curitiba usando técnicas de mineração de dados

Abstract

In this study, I analysed the possibility to apply data mining to the public database containing data provided by the health TI system of the city of Curitiba, the ESaúde.The E-Saúde contains data related to the medical care provided by the city. I applied data mining techniques, within a Knowledge Discovery Process, in the instances of the E-Saúde database, to study the possibility to find patterns and correlations, in the form of textual rules that could be used by managers of the Health Units to take decisions. In order to reach this objective, the Knowledge Discovery in Databases (KDD) process was applied to the database. However, because the characteristics of the database, it was necessary to adapt the KDD. The E-Saúde files made available by Curitiba's Open Data Portal were imported into a MySQL database, in which they were cleaned, processed, selected and validated. For the data mining phase, the classification task was applied using two algorithms, a Decision Tree inductor and another based on Rules. The algorithms were applied on all instances of the first quarter of 2017 in two city Districts. In the experiments I used the Weka, a free software, which contains a collection of algorithms ready for use. The Decision Tree algorithm used was J48, and the algorithm based on Rules applied was the JRip. The J48 and JRip classifiers were applied in the attribute Solicitação de Exames. This attribute is one of those that best define the resolutivity of a medical appointment. The resolutivity is related to if the patient had (or had not) his case solved. The obtained models can be considered valid, since the accuracy obtained were around 74%, indicating the applicability of the data mining process in the E-Saúde database. The results indicate that the E-Saúde can be explored to obtain knowledge potentially useful for the decision make process preformed by the managers of Basic Health Units.