Feature Selection for Classification of Hyperspectral Data by SVM

By Mahesh Pal, Giles M. Foody. IEEE Transactions on Geoscience and Remote Sensing, 2010

831 citations recorded by the source corpus.

Support vector machines (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, therefore, does not require a dimensionality-reduction analysis in preprocessing. Here, a series of classification analyses with two hyperspectral sensor data sets reveals that the accuracy of a classification by an 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, particularly if a small training sample is used. This highlights a dependence of the accuracy of classification by an SVM on the dimensionality of the data and, therefore, 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 noninferior (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 preprocessing operations for classification by an SVM.

View Paper: https://www.semanticscholar.org/paper/726ee80d13758113245d8ed14abcc9ca490736d0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Deep One-Class Classification (2018, background): “Apart from the necessity to perform explicit feature engineering (Pal & Foody, 2010), another drawback of the aforementioned methods is their poor computational scaling due to the construction and manipulation of the kernel matrix.”
An SVM Ensemble Approach Combining Spectral, Structural, and Semantic Features for the Classification of High-Resolution Remotely Sensed Imagery (2013, methodology): “However, a recent study shows that the classification accuracy by an SVM varies as a function of the number of features used, and the accuracy may decline significantly with the addition of features [35].”
A Review of Feature Selection Methods for Machine Learning-Based Disease Risk Prediction (2022, background): “Furthermore, the SNPs that are incorporated into the predictive model (following feature selection) are typically assumed to be associated with loci that are mechanistically or functionally related to the underlying disease etiology (Pal and Foody, 2010; López et al., 2018).”
Tree Species Classification in Boreal Forests With Hyperspectral Data (2013, background): “In addition to the choice of classifier, in the presence of hyperdimensional feature spaces, a feature selection/reduction process is important [23].”
Random Forest classification of Mediterranean land cover using multi-seasonal imagery and multi-seasonal texture (2012, methodology): “More recent methods are wrappers which include the use of non-parametric algorithms like classification trees, neural networks and support vector machines (Bazi and Melgani, 2006; De Stefano et al., 2007; Del Frate et al., 2005; Pal and Foody, 2010; Yu et al., 2002).”
SVM-Based Fuzzy Decision Trees for Classification of High Spatial Resolution Remote Sensing Images (2012, background): “relevance [57], and the random-forest-based FS [35],[58].”
Implementation of machine-learning classification in remote sensing: an applied review (2018, methodology): “For SVM, low values for the cost ( C ) parameter greatly decreased the kappa statistic; however, the performance stabilizes as C increases.”
Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework (2018, methodology): “Among these methods, SVMs are the most commonly used classifiers for HSI classification, because SVMs perform robustly with high-dimensional input data [13], [14].”
An advanced ACO algorithm for feature subset selection (2015, background): “Often, however, features are highly correlated, and this can suggest a degree of redundancy in the available information which may have a negative impact on classification accuracy [28].”
Recent advances and emerging challenges of feature selection in the context of big data (2015, background): “Given its ability to enhance the performance of learning algorithms, feature selection has attracted increasing interest in the field of machine learning, in processes such as clustering [17,18], regression [19,20] and classification [12,21], whether supervised or unsupervised.”