Coordenadas sintéticas em bancos de dados confidenciais: uma aplicação em dados de covid-19

Abstract

Many data collected by agencies have confidential characteristics and sensitive information, so research institutions must obey legal and ethical protocols not to disclose such information indiscriminately. This work uses the methodology of synthetic data and multiple imputation, which are techniques developed for the safe disclosure of sensitive data, since they present a greater preservation of the usefulness of the data. This method replaces the original values with simulated values using probability distributions fitted to the original values, and can be applied to replace partially or completely the original data. The model by [26] and updated by [25], uses this methodology to generate synthetic geographic coordinates, however the model did not include the prediction of non-inhabitable spaces, such as airports and lakes. Therefore, we contribute to the inclusion of such spaces and call them restricted areas (spaces where individuals do not live). To evaluate this contribution in the model, we used a simulated database and graphically represented the results of the application with and without the inclusion of restricted areas. Finally, we carried out the application in a database of COVID-19 cases in the city of Montes Claros - MG, and we were able to prove the importance of including uninhabitable spaces in the data for the generation of synthetic coordinates.