| Type: | Package |
| Title: | Sequential Forward Floating Selection using Jeffries-Matusita Distance |
| Version: | 0.2 |
| Date: | 2020-12-28 |
| Author: | Michele Dalponte and Hans Ole Oerka |
| Maintainer: | Michele Dalponte <michele.dalponte@fmach.it> |
| Description: | Feature selection using Sequential Forward Floating feature Selection and Jeffries-Matusita distance. It returns a suboptimal set of features to use for image classification. Reference: Dalponte, M., Oerka, H.O., Gobakken, T., Gianelle, D. & Naesset, E. (2013). Tree Species Classification in Boreal Forests With Hyperspectral Data. IEEE Transactions on Geoscience and Remote Sensing, 51, 2632-2645, <doi:10.1109/TGRS.2012.2216272>. |
| License: | GPL-3 |
| LazyData: | TRUE |
| RoxygenNote: | 7.1.1 |
| NeedsCompilation: | no |
| Packaged: | 2020-12-28 12:23:12 UTC; dalpontem |
| Repository: | CRAN |
| Date/Publication: | 2021-01-07 13:50:02 UTC |
Bhattacharyya distance among classes
Description
Bhattacharyya distance.
Usage
BHATdist(g, X)
Arguments
g |
A column vector of the lables. length(g) is equal to nrow(X). |
X |
A dataframe of the features. ncol(X) is equal to the total number of features, and nrow(X) is equal to the number of avaialble training samples. nrow(X) is equal to length(g) |
Value
A list containing a matrix of the class combinations and a vector of the Bhattacharyya distances of all the class combinations.
Author(s)
Michele Dalponte and Hans Ole Oerka
References
Dalponte, M., Oerka, H.O., Gobakken, T., Gianelle, D. & Naesset, E. (2013). Tree Species Classification in Boreal Forests With Hyperspectral Data. IEEE Transactions on Geoscience and Remote Sensing, 51, 2632-2645.
Jeffries-Matusita distance among classes
Description
Jeffries-Matusita distance.
Usage
JMdist(g, X)
Arguments
g |
A column vector of the lables. length(g) is equal to nrow(X). |
X |
A dataframe of the features. ncol(X) is equal to the total number of features, and nrow(X) is equal to the number of avaialble training samples. nrow(X) is equal to length(g) |
Value
A list containing a matrix of the class combinations and a vector of the JM distances of all the class combinations.
Author(s)
Michele Dalponte and Hans Ole Oerka
References
Dalponte, M., Oerka, H.O., Gobakken, T., Gianelle, D. & Naesset, E. (2013). Tree Species Classification in Boreal Forests With Hyperspectral Data. IEEE Transactions on Geoscience and Remote Sensing, 51, 2632-2645.
Hyperspectral data acquired over a forest area
Description
A dataset containing 3230 samples with 65 hyperspectral bands and 8 classes.
Usage
data(dat)
Format
A data frame with 3230 rows and 66 variables
Details
B1...B65 Hyperspectral bands.
SP. Classes.
Sequential Forward Floating Selection using Jeffries-Matusita Distance
Description
Feature selection using the Sequential Forward Floating Selection search strategy and the Jeffries-Matusita distance.
Usage
varSelSFFS(g, X, strategy = "mean", n = ncol(X))
Arguments
g |
A column vector of the lables. length(g) is equal to nrow(X). |
X |
A dataframe of the features. ncol(X) is equal to the total number of features, and nrow(X) is equal to the number of avaialble training samples. nrow(X) is equal to length(g) |
strategy |
string indicating the multiclass strategy to adopt: 'minimum' or 'mean'. |
n |
integer indicating the number of features to select. The algorithm will stop at n+1 features selected. |
Value
A list containing a vector of the JM distances on the individual bands, a matrix with the set of features selected, and a vector containing the distances for each feature set from 1 to N-1, where N is equal to ncol(X).
Author(s)
Michele Dalponte and Hans Ole Oerka
References
Dalponte, M., Oerka, H.O., Gobakken, T., Gianelle, D. & Naesset, E. (2013). Tree Species Classification in Boreal Forests With Hyperspectral Data. IEEE Transactions on Geoscience and Remote Sensing, 51, 2632-2645.
Examples
## Not run:
data(dat)
se<-varSelSFFS(g=dat$SP,X=dat[,c(1:65)],strategy="mean",n=4)
summary(se)
## End(Not run)