Annisa Fadhillah Pulungan, Desilia Selvida, Agnes Irene Silitonga
Early detection and prediction of HIV/AIDS infection are important challenges in the medical world. Early detection of HIV/AIDS infection is crucial for improving patients' quality of life and preventing the spread of the disease. However, the imbalances and anomalies in health data often reduce the performance of predictive models. This study proposes a combined approach using the Local Outlier Factor (LOF) algorithm for anomaly detection and Extreme Gradient Boosting (XGBoost) for predicting HIV infection status. The HIV/AIDS dataset used was obtained from the Kaggle platform with an imbalanced distribution (521 positive and 1,618 negative cases). LOF successfully filtered out 194 anomalous data points. The evaluation results showed that the combination of LOF+XGBoost significantly improved performance compared to pure XGBoost, with precision of 90.4%, recall of 90.5%, F1-score of 90%, and AUC of 0.92. The use of LOF has proven to enhance the sensitivity and balance of predictions, particularly when dealing with imbalanced data in the HIV/AIDS dataset. The combination of LOF+XGBoost significantly improves model performance in HIV disease detection. © 2025 IEEE.
Universitas Sumatera Utara, Faculty of Computer Science and Information Technology, Department of Information Technology, Medan, Indonesia; Universitas Sumatera Utara, Faculty of Computer Science and Information Technology, Department of Computer Science, Medan, Indonesia; Universitas Negeri Medan, Faculty of Economic, Department of Digital Business, Medan, Indonesia
