Prediction of Sustainability Related Properties: Data Science Methods with Incorporated Prior Knowledge
Book chapter, 2020

Many of the registered chemicals, newly synthesized or long existing, lack information on their hazard for the environment and human health. To perform the holistic safety assessment of chemicals, data on molecular properties, mostly obtained by test on living organisms, are collected. However, extensive experimentation is neither economically feasible nor ethical, and thus development of accurate prediction models is required. The recent advances in the area are associated with data science methods; however, there are certain limitations of these models with respect to their transparency, interpretability or even availability of well distributed training data to ensure robust generalization. Hybrid models combining machine learning with prior knowledge of the research field can potentially provide the solution to these limitations. The current study presents the first step of creating hybrid models, namely extraction of knowledge that can be utilized to create prior knowledge for future incorporation into hybrid models.

text mining

prior knowledge

hybrid model

Author

Gulnara Shavalieva

Energy Technology 3

Pietro Postacchini

RWTH Aachen University

Stavros Papadokonstantakis

Chalmers, Space, Earth and Environment, Energy Technology

Computer Aided Chemical Engineering

1570-7946 (ISSN)

1897-1902

Subject Categories

Other Computer and Information Science

Other Engineering and Technologies not elsewhere specified

Bioinformatics (Computational Biology)

DOI

10.1016/B978-0-12-823377-1.50317-7

More information

Latest update

3/21/2023