Publikation
Substitution of hazardous chemical substances using Deep Learning and t-SNE
Patrick Lübbecke; Nijat Mehdiyev; Peter Fettke
In: Proceedings der Internationalen Tagung Wirtschaftsinformatik. Internationale Tagung Wirtschaftsinformatik (WI-2019), Human Practice. Digital Ecologies. Our Future. February 24-27, Siegen, Germany, AIS, 2019.
Zusammenfassung
Manufacturing companies in the European Union are obliged to regularly analyze their recipes to find safer alternatives for hazardous substances. Unfortunately, available substance information is dispersed, heterogeneous and stored in databases of many private and public entities. In addition, the number of existing chemical substances already surpassed 85,000 with over 200 attributes for each substance, what makes it impossible for experts to collect and manually review this data. We tackle these issues by introducing a novel machine learning approach for alternative assessment. After
developing a central database, we design an approach which performs nearest neighbor search in latent space obtained by deep autoencoders. Furthermore, we implement a post-hoc explanation technique, t-SNE, to visualize deep embeddings that enables to justify model outcomes. The application in a real-world project with a manufacturer shows that this approach can help process experts to identify possible replacement candidates more quickly and fosters comprehensibility through visualization.