Reliability Improvement of Odour Detection Thresholds Bibliographic Data

Pascale Montreer (1) , Stefan Janaqi (2) , Stéphane Cariou (1) , Mathilde Chaignaud (3) , Isabelle Betremieux (4) , Philippe Ricoux (4) , Frédéric Picard (5) , Sabine Sirol , Budagwa Assumani , Jean-Louis Fanlo (1, 3)

1 LGEI - Laboratoire de Génie de l'Environnement Industriel
2 LGI2P - Laboratoire de Génie Informatique et Ingénierie de Production
3 Olentica
4 TOTAL S.A.
5 Hutchinson

Odour control is an important industrial and economical issue as it is a criterion in purchase and use of a material. The minimal concentration of a pure compound allowing to perceive its odour, called Odour Detection Threshold (ODT), is a key parameter of the odour control. Each compound has its own ODT. Bibliographic data are the main source of information to obtain ODTs. Nevertheless, there are a lot of not documented compounds and available ODTs are marred by a high variability. Another expensive and time-consuming way to obtain ODT is the measurement. This paper proposes a cleaning methodology to reduce available ODTs data uncertainty. This methodology will be consolidated by our own measurements. Next, we predict missing ODTs as a function of chemical and physical certain variables. The cleaning proposed leads to eliminate 39% of compounds with at least one ODT and 84% of positive cases (with a comparison on 37 compounds). The missing ODTs are predicted with an error of 0.83 for the train and 1.14 for the test (on a log10 scale). Given the uncertainty of data, the model is sufficient. This approach leads to work with a lower uncertainty on available ODTs and predict missing ODTs with a satisfactory model.

Mots clés: Odour Detection Thresholds (ODT); Data mining Reliability

Mots clés: Odour Detection Thresholds (ODT)

Domaines: Apprentissage [cs.LG]

Pascale Montreer, Stefan Janaqi, Stéphane Cariou, Mathilde Chaignaud, Isabelle Betremieux, et al.. Reliability Improvement of Odour Detection Thresholds Bibliographic Data. IPMU 2018 17th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems., Jun 2018, Cadiz, Spain. pp.562-573, ⟨10.1007/978-3-319-91473-2_48⟩. ⟨hal-01955230v2⟩