Knowledge discovery for health management: Application to COVID-19 data.

Abstract

Organizations today have increasingly large and complex data sets. Knowledge discovery in databases (KDD) and data mining techniques are used to find the required information, discover novel patterns and perform categorization. In particular, hospital databases have a great potential to explore hidden patterns in medical domain datasets due to their voluminous, heterogeneous and distributed nature. These patterns can be used for clinical diagnosis and resource management, among others. Given the recent pandemic situation, knowledge discovery for efficient decision making becomes imperative. This paper presents the use of combined k-means clustering techniques, k-prototypes and SOM self-organizing maps of the unsupervised competitive paradigm, on the public database with COVID-19 cases at the national level, from the Ministry of Health of Argentina. The results of k-prototypes allowed obtaining an overview of the distribution of samples, while the SOM, with high quality model evaluation measures, enabled a more complete, in-depth and visual analysis. In addition, software is presented to facilitate the study of results by experts.