Quantitative structure-property relationships (QSPR) have been determined to predict the melting point temperatures of 1,2,3-diazaborine compounds (n = 72). Electronic and topological descriptors were computed from molecular structures, and a QSPR model was generated by linear multiple regression using reported melting point temperatures as the dependent variable. The most important molecular descriptors describing this physicochemical property were the sum of the atomic charges for the heteroatoms, the sum of Randić connectivity indexes (0)Chi(0), (0)Chi(1), and (0)Chi(2), the total number of atoms in the molecule, and the volume of the box in which the molecule fits. The multiple determination coefficient (R(2)) and standard error of estimation for the model were 0.856 and 16.787 degrees C, respectively. In addition to regression techniques, a back-propagation neural network was used to include nonlinear relationships between molecular structure and melting point temperatures. It is concluded that melting point temperatures for 1,2,3-diazaborine compounds can be described by electrostatic interactions mediated by atomic charges and steric properties. The results of this study demonstrate that multiple linear regression analysis and back-propagation neural network are techniques that can be used to successfully predict the melting point temperatures of 1,2,3-diazaborine compounds. The most accurate prediction results were obtained using the Levenberg-Marquardt neural network algorithm.