The scientific design and preparation of porous carbon with high VOCs (such as acetone) adsorption capacity are crucial for waste gas treatment. However, it is still challenging to explain the microscopic mechanism of acetone uptake on porous carbon because of the different adsorption conditions, as well as various chemical properties and different pore structures of porous carbon. Here, the prediction model of the adsorption conditions, nitrogen/oxygen groups and pore structures on acetone adsorption capacity was established by machine learning, and the contribution of the nitrogen/oxygen groups and pore structures of porous carbons to the acetone adsorption capacity at different adsorption pressures was discussed. The results display that the nitrogen/oxygen groups and micropore volume are the main factors affecting the adsorption capacity of acetone at relatively low pressure, while the adsorption capacity at relatively high pressure is determined by total pore volume. Subsequently, three kinds of porous carbon with different oxygen content and gradient pore size distribution were synthesized, and acetone adsorption isotherms were tested. The effects of chemical properties and pore structure on acetone adsorption at different pressure were studied, and the results were consistent with those of machine learning. However, the results are challenging to illustrate the effect of single pore size and functional group type on acetone adsorption performance. Finally, molecular simulation was used to calculate acetone adsorption isotherms with various nitrogen/oxygen groups and different pore sizes, further revealing the adsorption mechanism of acetone with nitrogen/oxygen groups and pore size. Based on machine learning and molecular simulation results, new insights into the adsorption behavior of acetone were revealed, providing theoretical support for the sustainable development of carbon-based adsorbents for acetone waste gas treatment.