In this work a new topological index (TI) was proposed based on atom characteristics (e.g., atom radius, atom electronegativity, etc.) and atom positions in the hydrogen-suppressed molecule structure. Using the TIs, a multiple linear regression (MLR) model was developed for predicting the decomposition temperature (T-d) of 158 ionic liquids (ILs), which are based on imidazolium, pyridinium, pyrrolidinium, ammonium, phosphonium, sulfonium, and guanidinium. Because ILs are a class of molten salts which are composed entirely of cations and anions, in general, the descriptors for ILs are calculated from cations and anions separately, and the interaction between them is neglected. In this study, except for the two sets of TIs generated from cations and anions, a third TI was proposed to depict the interaction of anions and cations. The regression coefficient (R-2) and the overall average absolute deviation (AAD) are 0.893 and 3.07 %, respectively.