A data set of 412 olfactory compounds, divided into animal, camphoraceous, ethereal and fatty olfaction classes, was submitted to an analysis by a Fuzzy Logic procedure called Adaptive Fuzzy Partition (AFP). This method aims to establish molecular descriptor/chemical activity relationships by dynamically dividing the descriptor space into a set of fuzzily partitioned subspaces. The ability of these AFP models to classify the four olfactory notes was validated after dividing the data set compounds into training and test sets, including 310 and 102 molecules, respectively. The main olfactory note was correctly predicted for 83 % of the test set compounds.
Pintore, M., Audouze, K., Ros, F., & Chrétien, J. (2006). Adaptive fuzzy partition in database mining: application to olfaction. Data Science Journal, 1, 99–110. https://doi.org/10.2481/dsj.1.99