cpVisualise: Visual Interface for Unsupervised Changepoint Penalty...

In OllieFord/ChangepointVis: Interactive Visualization for Change Point Exploration and Labeling

Description

This function calculates the optimal positioning and (potentially) number of changepoints for a given univariate dataset using the user specified range of penalty values. It uses cpt.mean with the "PELT" method and "CROPS" penalty type from the changepoint package.

The function starts a shiny server and acompanying web interface that displays:

1. The univariate datset (top left)

2. General statistics (top right)

3. Solution path (bottom left)

4. Weighted means histogram of the segments between changepoints.

The main form of interaction with the interface is through selecting each penalty value and then inspecting the resulting solution on the main plot (top left).

Usage

cpVisualise(data, penalty_range = c(1e-5,10))

Arguments

data	A univariate dataset. This should be a vector containing the data within which you wish to find a changepoint.
penalty_range	A range of penalty values. This will vary based on the datset and is used to compute changepoints for the data for each penalty value.

Details

This function is used to find changes in mean for a given dataset using the PELT method. The function visualises the data, along side an interactive solution path, allowing the usert to inspect how different penalty values impact the resulting changepoint locations. Additional information is also presented to help the user interpret the data and specific penalty value. A table containing general statistics such as mean, variance etc and a weighted histogram, showing the distribution of segment means (weighted by length) for a particular penalty value.

For an extended usage example, please see the accompanying website: https://sites.google.com/view/changepointvis/home

Value

Starts a shiny app in a new window.

Saves the currently selected solution as a .csv file, for use in cpLabel.

Author(s)

Oliver Ford

References

PELT Algorithm: Killick R, Fearnhead P, Eckley IA (2012) Optimal detection of changepoints with a linear computational cost, JASA 107(500), 1590–1598
CROPS: Haynes K, Eckley IA, Fearnhead P (2014) Efficient penalty search for multiple changepoint problems (in submission), arXiv:1412.3617

Examples

## Not run: 
data = c(rnorm(100,0,1),rnorm(100,5,1))
penalty_range = c(1e-5,10)
cpVisualise(data, penalty_range)

## End(Not run)

OllieFord/ChangepointVis documentation built on Aug. 26, 2019, 7:53 p.m.