Prediction of critical micelle concentration for per- and polyfluoroalkyl substances

Numerical Approaches to Determine Cetane Number of Hydrocarbons and Oxygenated Compounds, Mixtures, and their Blends

“…Since the ML-based models proposed by our group in the early 2010s, , methodologies have evolved and numerous works have been published on the subject, and as mentioned above, new CN data have been made available. The CN prediction for pure compounds was performed by applying our now well-proven methodology for SVR-based model training, ,, to the data collections described above. Two cases were investigated: in the first, only the data set including pure compounds is taken into account resulting in type (I) models, while in the second, both pure compounds and mixtures are used to train models resulting in type (II) models.…”

“…From comparisons performed in previous studies, − each molecule in the pure compound database was encoded using descriptorslabeled as functional group count descriptors (FGCD)calculated on the basis of the chemical and structural formulas. In the FGCD family of molecular descriptors are included counts of atoms and groups of atoms identified as relevant from chemical aspects.…”

“…Since our first developments of QSPR based models for fuels property specifications , to very recent works on per- and polyfluoroalkyl substances, several ML algorithms have been tested in conjunction with different molecular structure encodings. It was widely demonstrated that the combination of SVM and FGCD provides accurate solutions in terms of property modeling. ,, The LibSVM library was used for Support Vector Regression (SVR) with both linear and radial basis function kernels, and with an epsilon insensitive zone.…”