Predictive analysis using gaussian processes regression and Mahalanobis distance approach: Anticipation of COVID-19 spike in Bandung City

Abstract

This paper presents research on the use of Gaussian process regression integrated with Mahalanobis distance to model the number of active COVID-19 cases, healed rates, and deaths from COVID-19 in Bandung City, Indonesia. Gaussian process regression as a machine learning method is used to predict the number of COVID-19 cases, while Mahalanobis distance is used to determine outliers, which are values that exceed a safe threshold as an early warning system. This analysis is important to anticipate possible spikes in cases that could occur in the future. Analysis was also carried out using a Pareto diagram to find out which region experienced the biggest spike. In addition, this study also considers the availability of beds in hospitals as health facilities that support the handling of COVID-19. The analysis results show that both analysis methods can model the COVID-19 case well. These findings can serve as a basis for policy-making related to COVID-19 handling.