Unsupervised machine learning predicts future sexual behaviour and sexually transmitted infections among HIV-positive men who have sex with men. PLoS Computational Biology
Machine learning is increasingly introduced into medical fields, yet there is limited evidence for its benefit over more commonly used statistical methods in epidemiological studies.
In the current study, Andresen et al. designed and tested a machine learning method, which was used to predict sexually transmitted diseases (STD’s) among HIV-positive men who have sex with men in Switzerland. They used the machine learning algorithm to find groups of men with similar sexual behaviour in the last twenty years. The main finding of the study was that considering these groups in addition to conventional risk factors yielded more accurate predictions of who would be diagnosed with an STD.
In summary, the study found that clustering individuals based on their sexual behaviour over time using unsupervised machine learning identifies subgroups of the MSM population in the SHCS with distinct STD exposure. These findings have implications for research on sexual health and STIs by proposing an alternative method for creating exposure categories for infectious disease modelling. Furthermore, this study contributes to the evidence on machine learning applications in epidemiology by validating a flexible unsupervised machine learning framework adaptable to any type of longitudinal data.