Rapid Evaluation of Prediction Methods with DIPPR’s Automated Property Prediction Package

An automated property prediction package has been developed that permits rapid evaluation of group-contribution, corresponding states, empirical, and theoretical property estimation methods. The property prediction package, which is part of the DIPPR^® Information And Data Evaluation Manager (DIADEM) software, is used in conjunction with the DIPPR^® 801 database to develop and test new prediction methods. The software is freely available to all DIPPR sponsor companies, but is also commercially available. The estimation engine is based on an automated SMILES (Simplified Molecular Input Line Entry Specification) formula parser to provide required molecular structural information, retrieval of required secondary properties from the DIPPR^® database, and defined rules for the method. Automatic comparisons of predicted values to experimental data in the DIPPR^® database can be made for properties at specified accuracy levels, by chemical family or type, or over the entire database. This allows evaluation of the relative effectiveness of methods for specific chemical families and tailoring of the selected method to specific chemical classes. New methods can readily be added by input using a simple input form. Nearly 200 thermophysical property prediction methods are currently available in DIADEM.