Greenhouse gas (GHG) emissions, which are closely related to climate change and serious ecological instability, have attracted global attention. The estimation of crucial aquatic factors for the flux of GHGs in lakes is a key step in controlling and reducing GHG emissions. The importance of 14 aquatic factors for GHG emissions was estimated in Meiliang Bay, which is an eutrophication shallow bay in Taihu Lake in eastern China. The random forest (RF) method, which is an improved version of the classified and regression tree (CART) model, was employed. No distribution assumption on variables was required in this method and it could include nonlinear actions and interactions among factors. The results show significant positive correlations among the fluxes of CO2, CH4, and N2O. The most crucial factor influencing CO2 emissions is the water temperature (WT) followed by sulfate (SO42−), alkalinity (Alk), dissolved oxygen (DO), and nitrate (NO3−–N). The important factors for CH4 emissions are WT, SO42−, DO, Alk, and NO2−–N. The outcome for N2O, in which the key factor is NO2−–N, was slightly different from those of CO2 and CH4. A comprehensive ranking index (CRI) for the fluxes of all three GHGs was also calculated and showed that WT, NO2−–N, SO42−, DO, and Alk are the most crucial aquatic factors. These results indicate that increasing DO might be the most effective means of controlling GHG emissions in eutrophication lake bays. The role of SO42− in GHG emissions, which has previously been ignored, is also worth paying attention to. This study provides a useful basis for controlling GHG emissions in eutrophication shallow lake bays.