Doctoral thesis
Open access

Enhancing HIV Prediction through Machine Learning and Streamlining Research with LiteRev

ContributorsOrel, Erol
Number of pages50
Imprimatur date2023
Defense date2023

High-yield HIV testing strategies are essential for controlling the epidemic in high prevalence, low-resource settings in East and Southern Africa. Similarly, literature reviews are crucial in synthesising relevant research to advance understanding and support decision-making. However, traditional HIV testing methods may not be sufficiently effective in identifying the remaining individuals who are unaware of their HIV-positive status. and traditional literature reviews are time-consuming and quickly outdated.

This thesis combines two significant advancements in public health research: the adoption of highly precise HIV testing methods in East and Southern Africa, and the creation of LiteRev, an automated tool designed to quicken and simplify the literature review process. Both of these developments utilise modern machine learning technologies.

One facet of this thesis focused on predicting the HIV status of individuals in ten African countries using a minimal set of socio-behavioural characteristics. Four algorithms (penalised logistic regression, generalised additive model, support vector machine, and gradient boosting trees) were employed for this analysis. The gradient boosting trees algorithm demonstrated superior performance in predicting HIV status with high precision, underscoring its potential in informing targeted HIV testing strategies and interventions.

Concurrently, LiteRev was applied to the domain of HIV research, specifically examining the burden and care for acute and early HIV infection in sub-Saharan Africa. LiteRev's performance, characterised by its ability to process and synthesise literature efficiently, was compared against traditional manual methods. It exhibited its ability to significantly reduce the workload involved in screening through many research papers.

The integration of advanced computational methods in public health research, as demonstrated by the predictive algorithms for HIV status and LiteRev, showcases the potential of technology in enhancing research efficiency and precision in critical health interventions. These tools not only provide valuable insights for developing targeted health strategies in challenging environments like sub-Saharan Africa but also accelerate and enrich literature reviews.

  • Machine Learning
  • HIV
  • Literature Review
Citation (ISO format)
OREL, Erol. Enhancing HIV Prediction through Machine Learning and Streamlining Research with LiteRev. 2023. doi: 10.13097/archive-ouverte/unige:175448
Main files (1)
Secondary files (1)

Technical informations

Creation06.03.2024 11:54:51
First validation06.03.2024 13:49:26
Update time06.03.2024 13:49:26
Status update06.03.2024 13:49:26
Last indexation06.05.2024 18:06:43
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack