QSPR Modeling of the Dielectric Constant Based on Monte Carlo Optimization Method

Monte Carlo Optimization method has been used to develop QSPR models for the prediction of the dielectric constant of organic compounds. The QSPR models were developed from the dataset of 201 diverse compounds, splited in three random splits in training and test set and based on SMILES notation based molecular descriptors. By employing a variety of statistical methodologies, an assessment of the predictive capabilities and resilience of the established QSPR models was achieved. The demonstrated numerical values used for their validation underscore the strong suitability of these QSPR models. For the best model statistical quality is calculated with following characteristics r2 = 0.7667, q2 = 0.7597, s = 4.81 for training and r2 = 0.8156, q2 = 0.7973, s = 5.00 for test set. The Monte Carlo optimization technique effectively identified molecular fragments with defined impact on dielectric constant represented in QSPR modeling through the use of SMILES notation based descriptors.