In lvaudor/hubr: Function to Segment Quantitative Series through the Hubert-Kehagias Algorithm.

knitr::opts_chunk$set(echo = TRUE)
set.seed(33)

library(hubr)

Let us consider the following signal, which is composed of several segments with differing lengths and means (but constant variance):

x=c(rnorm(130,3,2),
    rnorm(220,10,2),
    rnorm(500,2,2),
    rnorm(50,-4,2),
    rnorm(30,2,2),
    rnorm(120,8,2))
plot(x, type="l")

The Hubert_segmentation() function detects automatically the breaks in this shifting means signal:

seg_obj=Hubert_segmentation(x)
plot(x, type="l")
lines(model_signal(x,seg_obj$locations), col="red",lwd=3)

model_signal() is a helper function which returns the series of predicted values according to the segmentation object.

Now, let us consider a signal with shifting means, in which the shift in mean is a bit less important when compared to variance.

x=c(rnorm(130,3,2),
    rnorm(220,5,2),
    rnorm(500,4,2),
    rnorm(50,6,2),
    rnorm(30,4,2),
    rnorm(120,3,2))
plot(x, type="l")

seg_obj=Hubert_segmentation(x)
plot(x, type="l")
lines(model_signal(x,seg_obj$locations), col="red",lwd=3)

lvaudor/hubr documentation built on May 3, 2019, 7:40 p.m.