Amir Varamesh - University of Tehran
Forough Ameli - Iran University of Science and Technology
Abdolhossein Hemmati-Sarapardeh - Amirkabir University of Technology

In this article, least square support vector machine (LSSVM) is used to develop a fast and accurate non-group contribution model for estimation of molecular weight and acentric factor of pure components. Experimental data of 563 pure compounds from various chemical groups are used to develop this predictive model. Using this robust model, excellent results are achieved in estimation of molecular weight and acentric factor. Average absolute percent relative error of 6.39% and 2.17% are obtained for prediction of molecular weight and acentric factor, respectively. Furthermore, statistical and graphical error analysis proved that the proposed model gives more accurate prediction in comparison to previously developed methods

: Molecular weight, acentric factor, non-group contribution, least square support vector machine