Find Cutting Points
Finds cutting points for vector s of real numbers.
cutpoints(x, nmax = 8, quant = 0.95)
x |
vector of real values. |
nmax |
the maximum number of cutting points to choose |
quant |
quantile of the gaps to consider for cuts. |
Finds cutting points for vector s of real numbers, based on the gaps in the values of the vector. The number of cutting points is derived from a quantile of gaps in the values. The user can set a lower limit for this number of gaps.
Returns a list with components cutp, the cutting points selected,
and cutd, the gap between values of x at this cutting point.
Automatically finding cutting points is often requested in Data Mining. If a target attribute is available, Quinlan's C5.0 does a very good job here. Unfortunately, the ‘C5.0’ package (of the R-Forge project “Rulebased Models”) is quite cumbersome to use.
Witten, I. H., and E. Frank (2005). Data Mining: Practical Machine Learning Tools and Techniques. Morgan Kaufmann Publishers, San Francisco.
N <- 100; x <- sort(runif(N))
cp <- cutpoints(x, 6, 0.9)
n <- length(cp$cutp)
# Print out
nocp <- rle(findInterval(x, c(-Inf, cp$cutp, Inf)))$lengths
cbind(c(-Inf, cp$cutp), c(cp$cutp, Inf), nocp)
# Define a factor from the cutting points
fx <- cut(x, breaks = c(-Inf, cp$cutp, Inf))
## Not run:
# Plot points and cutting points
plot(x, rep(0, N), col="gray", ann = FALSE)
points(cp$cutp, rep(0, n), pch="|", col=2)
# Compare with k-means clustering
km <- kmeans(x, n)
points(x, rep(0, N), col = km$cluster, pch = "+")
## A 2-dimensional example
x <- y <- c()
for (i in 1:9) {
for (j in 1:9) {
x <- c(x, i + rnorm(20, 0, 0.2))
y <- c(y, j + rnorm(20, 0, 0.2))
}
}
cpx <- cutpoints(x, 8, 0)
cpy <- cutpoints(y, 8, 0)
plot(x, y, pch = 18, col=rgb(0.5,0.5,0.5), axes=FALSE, ann=FALSE)
for (xi in cpx$cutp) abline(v=xi, col=2, lty=2)
for (yi in cpy$cutp) abline(h=yi, col=2, lty=2)
km <- kmeans(cbind(x, y), 81)
points(x, y, col=km$cluster)
## End(Not run)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.