Novel method for properties prediction of pure organic compounds using machine learning

Simple item page
dc.contributor.authorChorbngam, Nattasinee
dc.contributor.authorChawuthai, Rathachai
dc.contributor.authorAnantpinijwatna, Amata
dc.date.accessioned2026-08-06T10:31:15Z
dc.date.available2026-08-06T10:31:15Z
dc.date.issued2021-01-01
dc.description.abstractIn classical thermodynamic, the estimation method of pure compounds properties was based on Newtonian physics, which required experimental data. It is proven to be inadequate for the growing demand of the novel chemical synthesis. There were several studies on the prediction of the pure compound properties based on QSPR methods. However, the conventional group-contribution based methods predictive capability was limited by the available measured data. Therefore, this study aims to approach the property prediction with a novel statistical-based method. The proposed method is derived using supervised machine learning algorithms. The experimental data used to train and validate the models were collected from the published literature. These data set are composed of the alkanes, alkenes, and alkynes derivatives containing 1-12 carbon atoms. The results show the improved accuracy of the model prediction compare to the conventional method in terms of root mean square error (RMSE) and mean absolute percentage error (MAPE).
dc.identifier.citationComputer Aided Chemical Engineering, 50, 431-437, 2021
dc.identifier.doi10.1016/B978-0-323-88506-5.50068-1
dc.identifier.issn15707946
dc.identifier.other2-s2.0-85110283545
dc.identifier.urihttps://dspace.kmitl.ac.th/handle/123456789/11666
dc.sourceComputer Aided Chemical Engineering
dc.subjectMachine learning
dc.subjectOrganic compounds
dc.subjectProperties prediction
dc.titleNovel method for properties prediction of pure organic compounds using machine learning
dc.typeBook Chapter

Files

Collections

Research