Skip to content


This software allows to segment multi-dimensional time series into frequent, rare and extreme events in an unsupervised way (without any expertise required in the process). A modeling of the dynamics of these events is presented by a graph and its dynamics (probability of being in an event or not, probability of event changing) thanks to a Hidden Markov Model.

Two variants are available:

  • uHMM: a conventional R-package including a tcl/tk interface
  • uHMMweb: a demo web software written in R-Shiny

uHMM, a R package and its tcl/tk-interface

The uHMM interface and its R package are distributed on the CRAN website: here.

Installation Procedure

It can be directly installed from a R session. Open a R session, then execute the following instructions:


How to use it

In a R session, execute:


uHMMweb, an online demo version

The online app uHMMweb has the same features than uHMM. It can be launched here: uHMMweb.