Practical use of statistical learning theory for modeling freezing point depression of electrolyte solutions: LSSVM model

• LSSVM algorithm is utilized to calculate the FPD of different electrolyte solutions.
• The predictions of the model are compared to a rigorous thermodynamic model.
• Statistical and graphical error analyses are presented to demonstrate the accuracy.

Electrolyte solutions are mixtures comprising a substance with the capability of forming strong associating bonding interactions between molecules. Hence, the predictions of van der Waals based equations of state for properties of these systems are poor. In these cases, employment of an equation of state (EoS) combined with the association term from the statistical associating fluid theory (SAFT) has been recommended in the literature. In this communication, a robust type of learning method developed based on statistical learning theory namely least squares support vector machine (LSSVM) has been employed for calculating the freezing point depression (FPD) of different electrolyte solutions. The predictions of the developed model are compared to the results of cubic-plus-association (CPA) EoS combined with the Debye–Hückel electrostatic term. It is found that the proposed smart technique gives more accurate estimations than CPA EoS that enjoys SAFT for the association part.