MODELE PREDICTIF DES PRESTATIONS SOCIALES À L’AIDE DE LA FORÊT ALÉATOIRE : ÉTUDE DE CAS À LA CNSS RDC

Social security institutions in low- and middle-income countries operate in increasingly complex environments marked by demographic pressures, financial constraints and unstable labour markets. In the Democratic Republic of Congo, the National Social Security Fund (CNSS) manages large volumes of heterogeneous data related to contributions and benefit payments, yet most analyses remain descriptive. This study develops a Random Forest predictive framework using historical data stored in the CNSS data warehouse to forecast both the probability and amount of future benefits. The approach combines structured data preparation, feature engineering, model training and evaluation through metrics such as accuracy, precision, recall, F1-score, RMSE and R². Findings show strong predictive performance, with contribution regularity, benefit history, age and seniority emerging as key determinants. These results demonstrate the potential of machine learning to support evidence-based planning, financial forecasting and risk management within social security systems in the DRC.