<![CDATA[This study aims to analyze the performance of machine learning algorithms in classifying HIV-related risk based on opportunistic infection indicators in population-level aggregate data. The dataset was obtained from Kaggle and contains global mortality data, with the target variable constructed as a binary classification based on the median number of HIV/AIDS deaths to represent high- and low-risk categories at the population level, rather than individual diagnosis. The research stages include data preprocessing, handling class imbalance using SMOTE on the training data, feature selection based on clinical domain knowledge, and model training using Random Forest, Support Vector Machine (SVM), and Logistic Regression with GridSearchCV optimization and cross-validation. The results show that Random Forest achieved the best performance, with an accuracy of 98.56 percent and an AUC of 0.99. However, this performance should be interpreted cautiously because it is influenced by the high correlation among features and the median-based target construction, which may simplify the classification patterns. This study demonstrates that machine learning can be used to identify HIV-related risk patterns in population-level aggregate data. However, the resulting model is not intended for individual clinical diagnosis, but rather as a risk analysis tool at the population level.]]>
Copyrights © 2026
