Feature selection for classification of hyperspectral data by SVM
Foody, Giles M.
- Publisher: Institute of Electrical and Electronics Engineers
SVM are attractive for the classification of remotely sensed data with some claims that the method is insensitive to the dimensionality of the data and so not requiring a dimensionality reduction analysis in pre-processing. Here, a series of classification analyses with two hyperspectral sensor data sets reveal that the accuracy of a classification by a SVM does vary as a function of the number of features used. Critically, it is shown that the accuracy of a classification may decline significantly (at 0.05 level of statistical significance) with the addition of features, especially if a small training sample is used. This highlights a dependency of the accuracy of classification by a SVM on the dimensionality of the data and so the potential value of undertaking a feature selection analysis prior to classification. Additionally, it is demonstrated that even when a large training sample is available feature selection may still be useful. For example, the accuracy derived from the use of a small number of features may be non-inferior (at 0.05% level of significance) to that derived from the use of a larger feature set providing potential advantages in relation to issues such as data storage and computational processing costs. Feature selection may, therefore, be a valuable analysis to include in pre-processing operations for classification by a SVM.