“…In general, smaller or molecular family-specific (e.g., alcohols, alkanes, hydrocarbons, unsaturated hydrocarbons, and heterogeneous molecules) data sets result in lower errors for the boiling point, flash point, and yield sooting index models [11,12,16,17]. Melting point and heat of combustion models, however, have lower errors with the largest data sets [10,22].…”