Forecasting trends of HIV infection using deep learning models in East Gojjam zone, North West Ethiopia, 2025

BackgroundThe growing burden of HIV/AIDS, particularly in sub-Saharan Africa, presents a significant public health challenge, characterized by increasing morbidity, and mortality rates. This region is disproportionately affected, bearing for two-thirds of the global HIV/AIDS problem, highlighting an urgent need for effective solutions. Accurate forecasting of new HIV infections is crucial for developing targeted interventions to combat the HIV/AIDS pandemic. ObjectiveThis study aims to forecast trends of new HIV infections for the next five years and identify the contributing factors in the East Gojjam Zone. MethodsDHIS2 (2018-2025) data set from East Gojjam zone were analyzed using to a hybrid machine learning and deep learning framework. Machine learning models (Decision Tree, Random Forest, XGBoost, LightGBM, CatBoost, AdaBoost, and Gradient Boosting) were used for feature selection, and deep learning architectures (RNN, LSTM, GRU, and bidirectional variants) were used for time-series forecasting. Model performance was assessed using MAE, MSE, RMSE and MAPE ResultFrom the seven machine-learning algorithms used for selecting important futures the random forest was best performed model and many features were selected to apply for further forecasting using deep learning algorithms. Bidirectional LSTM model was best performed model among the six sequential deep learning algorithms used for forecasting HIV infection in East Gojjam zone. Forecasts reveal an upward trend of HIV infection in study area. ConclusionCombination of Machine learning and Deep learning algorithms method shows high predictive accuracy in forecasting of HIV infection. The forecasted trend shows an upward trend and needs urgent intervention and attention to combat the problem.