P1.0.21 Optimal Feature Selection for Classifying a Large Set of Chemicals Using Metal Oxide Sensors
- Event
- 14th International Meeting on Chemical Sensors - IMCS 2012
2012-05-20 - 2012-05-23
Nürnberg/Nuremberg, Germany - Chapter
- P1.0 Metal Oxide-based Sensors
- Author(s)
- T. Nowotny - School of Engineering and Informatics, University of Sussex (UK), A. Berna, S. Trowell - CSIRO Food Futures Flagship and Ecosystem Sciences Division (Australia), R. Binions - School of Engineering and Materials Science, Queen Mary University of London (UK)
- Pages
- 810 - 813
- DOI
- 10.5162/IMCS2012/P1.0.21
- ISBN
- 978-3-9813484-2-2
- Price
- free
Abstract
We investigated the feature selection problem for the application of all-against-all classification of a set of 20 chemicals using metal oxide sensors and linear support vector machines. We defined a set of possible features in terms of identity of sensors and sampling times and tested all possible combinations of such features. We found that performance is clearly increased by feature selection compared to previous results but that, contradictory to naïve expectation, using the maximal number of different sensors and all available data points for each sensor does not necessarily yield the best results. Similarly, the standard method of taking one data point from all sensors also underperforms.