Differential COVID-19 Vaccine Effectiveness and Predictive Survival Modeling: Data Mining Analysis in Peru

Abstract

The study aims to develop descriptive and predictive models of COVID-19 behavior in Perú, using data mining techniques to support public health policies based on epidemiological evidence. The CRISP-DM methodology was applied, using decision tree, clustering, and Naive Bayes algorithms on the national database “Deaths, hospitalizations, and vaccinations due to COVID-19,” processed in RapidMiner Studio. The results showed that, out of 8,120 cases of post-vaccination infection with three doses, 61 people died, with a fatality rate of 0.75%. The average age of those affected was 52 years. Pfizer doses were distributed as follows: first (73.7%), second (77.8%), and third (99%). Decision trees demonstrated superior predictive effectiveness, revealing a significant reduction in mortality correlated with the number of doses administered. These findings highlight the usefulness of predictive models for optimizing vaccination strategies in vulnerable populations.