“…QSPR models for DCN, RON, and MON differ in the way they encode the molecular structure. Various descriptors have been used including structural group counts (e.g., in [12][13][14]21,[29][30][31][32] ), the number of aromatic bonds (e.g., in 13,27 ), and topological indices, such as the Wiener Index 37 or branching indices (e.g., in 20,26,31 ). Previous models have also used a variety of techniques for the regression step, e.g., linear or nonlinear regression, 13,14,21,26,29,30 or artificial neural networks (ANNs).…”