Abstract: Sauce-flavor Xi Baijiu was a kind of sauce-flavor liquor with mellow flavor, elegant and long aftertaste, with various and complex aroma ingredients. In order to study the differences of the main volatile substances composition and aroma flavor composition of sauce-flavor Chinese Baijiu, a quantitative structure-retention relationship (QSRR) model of the GC retention index of aromatic ingredients in sauce-flavor Xi Baijiu was established. Moreover, a novel structure parameter-valence bond atomic index ^mA was derived based on molecular topology theory and the spatial and electrical structure of atoms in the aroma components molecules of sauce-flavor Xi Baijiu. The electrical topological state index (E_m) of aroma molecules of sauce-flavor Xi Baijiu was calculated, and ⁰A, E₁, E₂, E₁₄ were optimized screened. The four structure parameters were taken as input layer nodes, while the GC retention index was used as output layer nodes of the three-layer (4-12-1) structure, a neural network was constructed. The correlation coefficient was 0.9965. The mean relative error between the predicted value and the literature value of retention index RI was only 1.41%. The results showed that there was good non-linear relationship between the retention index, valence bond atomic index and electrical topological state index. Non-hydrogen atomic structure and –CH₃, >CH₂, –O– iso-group fragments were the main factors affecting the GC retention index of aromatic ingredients in sauce-flavor Xi Baijiu.