Predicting Antimicrobial Resistance Trends Combining Standard Linear Algebra with Machine Learning Algorithms

Abstract

Antimicrobial resistance prediction is a pivotal ongoing research activity that is currently being explored across various levels. In this context, we present the application of two prediction methods that model the antimicrobial resistance of Neisseria gonorrhoeae on the national level as an outcome of socio-economic processes. The methods use two different implementations of the principal component analysis combined with classification algorithms. Using these two methods, we generated forecasts concerning antimicrobial resistance of Neisseria gonorrhoeae, using publicly available databases encompassing over 200 countries from 1998 to 2021. Both approaches exhibit similar mean absolute averages and correlations when comparing available measurements with predictions. Steps of statistical analysis and applications are discussed, including population-weighted central tendencies, geographical correlations, time trends and error reduction possibilities.