Olfaction is one of the most complex biological properties to be processed, because:
the expert's subjectivity is involved in establishing an olfactory description
a given compound can exhibit simultaneously different odors
a very huge number of receptors are involved in the process
Standard data mining methods are not suitable to fulfil this objective.
Fuzzy logic mimics human reasoning in its ability to produce correct judgments from ambiguous and uncertain information. Then, it address the "fuzziness" linked to the expert's subjectivity in the odors characterization.
In a first work on 412 molecules subdivided into 4 olfactory notes, by a proprietary method named Adaptive Fuzzy Partition (AFP), we achieved a correct olfactory note prediction for 83% of the test set compounds. Later, on a larger set of 1,100 compounds and seven olfactory notes, we obtained an excellent score of about 80% of correct test prediction (unpublished results).
