Sensors practitioners don't make full use of the power of state-of-the-art pattern recognition (PR) algorithms and software. In this paper we apply -to our knowledge for the first time- Random Forests (RF) and Nearest Shrunken Centroids (NSC) to the classification of three E-Nose datasets of different hardness. We compare the classification rate with the one obtained by SVM. The classifiers parameters are optimized in an inner cross-validation (CV) cycle and the error is calculated by outer CV in order to avoid any bias. RF and SVM have a similar classification performance (SVM has an edge on the most difficult dataset). On the other hand, RF and NSC have an in-built feature selection mechanism that is very helpful for understanding the structure of the dataset and evaluating sensors.
Random Forests, Nearest Shrunken Centroids and support vector machines for the classification of diverse E-nose datasets
Pardo Matteo;
2006
Abstract
Sensors practitioners don't make full use of the power of state-of-the-art pattern recognition (PR) algorithms and software. In this paper we apply -to our knowledge for the first time- Random Forests (RF) and Nearest Shrunken Centroids (NSC) to the classification of three E-Nose datasets of different hardness. We compare the classification rate with the one obtained by SVM. The classifiers parameters are optimized in an inner cross-validation (CV) cycle and the error is calculated by outer CV in order to avoid any bias. RF and SVM have a similar classification performance (SVM has an edge on the most difficult dataset). On the other hand, RF and NSC have an in-built feature selection mechanism that is very helpful for understanding the structure of the dataset and evaluating sensors.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
