Fuzzy K-NN applied to moulds detection

The possibility to detect Aspergillus versicolor growing on different building materials by a metal oxide sensor array is studied. Results show that an accurate classification rate of 89 +/- 3% can be obtained combining an extended linear discriminant analysis plus a fuzzy k-NN classifier. The classification ability of the classifier is assessed within the dataset by crossvalidation and also in a second dataset collected 5 months later. There is a slight decrease in the classification performance for all the algorithms, being the most sensitive the most accurate one.