Análise de perfis de doenças com base em técnicas de descoberta de conhecimento em bases de dados

Abstract

This study aims to analyze disease profiles based on techniques for discovering the knowledge of databases of patients' medical records in the public network. The discovery of correlated information in medical records can be a complex and long-term task for health professionals, especially when the amount of data is large or without the aid of specialized computer systems. The process of knowledge discovery can be an alternative to obtain specialized knowledge from the interpretation of data and transform it into useful knowledge in the health area. For this purpose, a database of electronic medical records of users of the single health system was used, which has 43,879 patients and 2,296,626 patients in the city of Pato Branco, Paraná, in the year 2015. In this work, knowledge was discovered in order to find disease profiles using PostgreSQL tools for database use, j48, bagging and boosting algorithms with the WEKA tool. All technologies employed are of public use. The result was the acquisition and two groups of diseases, neoplasms that have 119 instances and trauma that in turn have 236 instances, for the application of the algorithms of data mining with a selection of 17 total attributes. Where the best performance was with running the Boosting algorithm to sort. The rules generated by the decision tree on the diseases of the group of neoplasms had main neighborhood branches, weight, school attendance, age group and height. Already for the group of diseases of injuries the main rules were about neighborhood, weight, height, age group, sex, diabetes and if they smoked. Based on the results it is possible to make an analysis of a specialist in the health area to confirm that the rules generated enable a new knowledge acquisition related to the profile of these diseases so that it is possible to have a greater prevention and complete the discovery of knowledge.