Detection of death patterns by COVID-19 using unsupervised data mining techniques

Fernández-Luque D.; Laura-Ochoa L.; Bedregal-Alpaca N. · RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao (2023)

Resumen

Data mining allows the discovery of new and meaningful relationships, patterns, and rules from large data sets, which support better decision-making. In this work, unsupervised data mining techniques are used to identify patterns of deaths registered in Peru due to COVID-19, using the K-means clustering algorithm and Apriori algorithm for association rules. Open data provided by the Ministry of Health (MINSA) of Peru were used, considering the period from January 2021 to April 2022. The phases of data understanding, data preparation, and modeling described in the CRISP-DM methodology were followed for subsequent pattern extraction and analysis of results. The results show an analysis of the cases of death according to the conditions of the patient prior to death, such as hospitalization conditions and vaccination dose. In addition, the highest risk profile for death is obtained in older male adults without vaccination doses. These results may help to have a vision of the current and future impact of the risk of death after infection and the effectiveness of vaccination in the country.

Instituciones: Universidad Nacional de San Agustín de Arequipa