Supervised Classification of Functional Data

Functions for classifying functional observations into discrete groups. Functional Classification

Usage

fclassif(
  fdataobj,
  y,
  method = c("lda", "qda", "knn", "kernel", "dd", "svm"),
  covariates = NULL,
  ncomp = 3,
  ...
)

Arguments

fdataobj

An object of class 'fdata'.

y

Integer vector of class labels (1-indexed).

method

Classification method: "lda", "qda", "knn", "kernel", "dd", or "svm".

covariates

Optional matrix of scalar covariates.

ncomp

Number of FPC components (default 3). Used by lda, qda, knn, svm.

...

Additional arguments:

k: Number of neighbors for kNN (default 5).
h.func: Bandwidth for functional kernel (default auto).
h.scalar: Bandwidth for scalar kernel (default auto).
kernel: SVM kernel type: "radial" (default), "linear", "polynomial", "sigmoid".
cost: SVM cost parameter (default 1).
gamma: SVM kernel parameter (default 1/ncomp).

Value

An object of class 'fclassif' with components:

predicted: Predicted class labels
probabilities: Posterior probabilities (if available)
accuracy: Training accuracy
confusion: Confusion matrix
method: Method used
ncomp: Number of FPC components

Details

Classifies functional data using one of several methods: LDA, QDA, kNN, kernel, DD-plot (depth-vs-depth), or SVM (support vector machine).

Examples

# \donttest{
fd <- fdata(matrix(rnorm(500), 50, 10), argvals = seq(0, 1, length.out = 10))
y <- rep(1:2, each = 25)
fit <- fclassif(fd, y, method = "lda", ncomp = 3)
fit$accuracy
#> [1] 0.54
# }