In a closed loop vapor compression cycle, a small portion of the oil circulates with the refrigerant flow through the cycle components while most of the oil stays inside the compressor. The worst scenario of oil circulation in the refrigeration cycle is when large amounts of oil become logged in the system. Each cycle component has different amounts of oil retention. Because oil retention in refrigeration systems can affect performance and compressor reliability, it receives continuous attention from manufactures and operators. In this paper, a group method of data handling (GMDH) type artificial neural network are used for modeling the effect of important parameters on oil retention in a carbon dioxide air-conditioning system is trained and tested with the experimental data taken from the experimental work by Lee [1].In this way, oil retention in a carbon dioxide air-conditioning system is modelled with respect to the variation of refrigerant mass flow rate, oil mass flow rate, oil circulation ratio, gas cooler inlet pressure, evaporator inlet pressure, gas cooler inlet temperature , gas cooler outlet temperature and evaporator outlet temperature which are defined as inputs.