pca: Feature extraction by principal component analysis
In kota7/MLPipe: Machine Learning Pipeline

Description Usage Arguments Value Class Methods Details Examples

Feature extraction by principal component analysis

1	pca_extractor(ncomp, center = TRUE, scale = TRUE)

ncomp: number of principal components to extract
center: either a logical value that indicates whether variables should be centered to zero-mean, or numeric vector of center values
scale: either a logical value that indicates whether variables should be scaled to unit-variance, or numeric vector of scale values

PCAExtractor class object

fit(x, y = NULL): conduct principal component analysis of x
transform(x, y = NULL): extract principal components of x
predict(x, y = NULL): returns all principal components of x

Uses prcomp as the backend.

p <- pca_extractor(2, center=TRUE, scale=TRUE)
p$fit(USArrests)
p$transform(USArrests)
p$predict(USArrests)
# inv_transform partially recovers the original data
cor(as.numeric(as.matrix(USArrests)),
    as.numeric(p$inv_transform(p$transform(USArrests)$x)$x))

p2 <-  pca_extractor(4, center=TRUE, scale=TRUE)
p2$fit(USArrests)
p2$transform(USArrests)
p2$predict(USArrests)
# when ncomp = ncol(x), inv_transform recovers the data perfectly
cor(as.numeric(as.matrix(USArrests)),
    as.numeric(p2$inv_transform(p2$transform(USArrests)$x)$x))