The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

scanr

scanr is an R package for sequential change-point detection in univariate long time series. The R interface calls a native Rust backend through extendr.

Installation

Users can install the package from CRAN and load it as follows:

install.packages("scanr")

Alternatively, install the development version from GitHub using pak:

install.packages("pak")
pak::pak("Prabashoka/scanr")
library(scanr)

Because the GitHub version is compiled from source, Rust and Cargo must be installed before installing scanr. To check whether they are available, run: Sys.which(c(“cargo”, “rustc”)) Both commands should return a valid file path. If they are not installed, follow the Rust installation instructions at https://rustup.rs/.

Basic Usage

The example below simulates a time series of length 20,000 with 20 change points in the mean. A scan window contains observations on each side of a candidate split. Smaller windows localize nearby changes more precisely but contain less information; larger windows are more stable but should remain smaller than the spacing between nearby changes.

set.seed(1234)

n <- 20000

change_points <- c(
  952, 1905, 2858, 3810, 4763, 5715, 6668, 7620, 8573, 9525,
  10478, 11430, 12383, 13335, 14288, 15240, 16193, 17145, 18098,
  19050
)

means <- c(
  0, 2, -1, 3, 0.5, -2, 2, 5, -0.5, 2.5, 0, -2.5, -1.5, 1.5,
  3, 1, 0, 1.25, -2, 3.5, -1.5
)

segment_starts <- c(1L, change_points + 1L)
segment_ends <- c(change_points, n)
x_mean <- numeric(n)

for (j in seq_along(means)) {
  segment_index <- segment_starts[j]:segment_ends[j]
  x_mean[segment_index] <- rnorm(length(segment_index), mean = means[j], sd = 1)
}

change_points

default_window_sizes() samples n_windows evenly spaced scales between the chosen bounds. Its upper bound defaults to floor(sqrt(n)) and cannot exceed floor(n / 2). Here we set problem-informed bounds because the simulated changes are roughly 950 observations apart. If window_sizes is omitted, scan_cpd() calls this helper using its min_window, max_window, and n_windows arguments.

window_sizes <- default_window_sizes(
  n = length(x_mean),
  min_window = 100,
  max_window = 737,
  n_windows = 11
)
window_sizes

fit_mean <- scan_cpd(
  x_mean,
  window_sizes = window_sizes,
  n_boot = 400,
  random_state = 1234,
  change_type = "mean",
  n_jobs = 1
)

fit_mean

Reference

Include the paper here.

License

This package is free and open source software, licensed under GPL-3.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.