Human Papillomavirus (HPV) is a major public health issue, particularly in low- and middle-income countries where vaccination coverage is low, healthcare access is limited, and there are barriers to effective screening. This study proposes a machine learning model that segments populations into risk profiles to better identify those at high risk for HPV infection and cervical cancer.

The model uses K-Means clustering to divide the population into groups based on demographic, behavioral, and medical data. Once these groups are identified, decision trees (specifically XGBoost) are used to predict the risk of HPV infection for each individual. This hybrid approach not only improves the prediction of HPV infection risk but also supports tailored public health interventions, such as prioritizing vaccination for high-risk individuals and subsidizing screening for those at medium risk.

By analyzing data from cervical cancer screenings at the University Hospital of Caracas, Venezuela, the model accurately identifies and categorizes individuals based on their likelihood of being infected. The study shows that this method has high accuracy, with a clear distinction between low, medium, and high-risk profiles, and can guide decision-making in resource-limited settings.

Additionally, the approach is adaptable to other diseases like HIV, making it a versatile tool for public health management.