Oxidative propane dehydrogenation with CO₂ (ODPC) is an economical and ecofriendly process that produces propylene and consumes CO₂ simultaneously. In this study, the catalyst composition for the ODPC reaction was optimized using a closed-loop optimization framework. A machine learning (ML) model was trained to predict the propylene yield and CO₂ conversion using an in-house experimental database obtained from metal oxide catalysts containing various elements. The trained ML model optimized the chemical composition of the catalysts and simultaneously maximized the propylene yield and CO₂ conversion using a metaheuristic algorithm. The proposed catalysts were prepared and their ODPC performance was evaluated. The data were included in the initial database to retrain the ML model. After this closed-loop optimization for 4 cycles, the proposed catalysts, which comprised four or five metal components, exhibited an enhanced ODPC performance compared with that of the initial database, which contained up to three metal components. Density functional theory calculations and characterization techniques were performed to investigate the role of each metal in the proposed catalyst. This study suggested a framework to optimize the chemical composition of multi-component catalysts to enhance the propylene yield and CO₂ activity in the ODPC reaction.