1. Group contribution model based on decision tree algorithm for estimating flash point of pure organic compounds.
- Author
-
Cao, Yan, Wang, Qiangfeng, Wang, Zhijie, and Ghadiri, Mahdi
- Abstract
One of the well-known chemical properties which has wide applications in the determination of the amount of flammability and combustibility of various compounds is flash point. Due to difficulty and high cost of experimental determination of flash points, computational and modeling studies of this parameter have been highlighted. In this work, a new model based on Ensemble Boosted Trees (EDT Boosted) is used to anticipate flash point of pure organic compounds as a function of normal boiling point and molecular structure. A comprehensive dataset of 1660 flash points is utilized to assess EDT Boosted model visually and mathematically. The mathematical indexes including R-squared, root mean square error, mean squared error, mean relative error, and standard deviations are forecasted 0.997, 4.5469, 20.6743, 0.85729, and 3.7123 respectively. Additionally, the visual comparison of EDT Boosted outputs and real values of flash point shows our proposed algorithm has great capacity in the calculation of flash point. Finally, the comparison of the suggested algorithm and other approaches is carried out to introduce EDT Boosted algorithm as the best predictor tool for flash point determination. [ABSTRACT FROM AUTHOR]
- Published
- 2020
- Full Text
- View/download PDF