Skip to content

Installation and Structure

A Graphical User Interface (GUI)

The uHMM2 interface is a graphical interface developed in R language [7] coupled with Tcl/Tk (help [8]), which enables the user to estimate a hidden Markov model built by unsupervised learning. The first version of this interface is available on the CRAN repository since 23 April 2016.

Installation of the interface

To install it, you just have to launch the following command in an R console :


## Installing package into '/home/hebert/programmation/R/packages'
## (as 'lib' is unspecified)
To launch the interface, use the following commands:

## <Tcl>

Structure of the interface

This interface is composed of 6 successive tabs:

  • the "Overview" tab (fig 1), which allows to choose a user mode (standard3 /expert4 ) and to change the language;
  • the "Import" tab, which allows you to manage files and data;
  • the "Variable selection" tab, which allows to select the variables of the model;
  • the "Classification" tab, which allows (in expert mode) to choose the state detection method;
  • the "Time series modeling" tab, which allows to estimate the parameters of the uHMM model;
  • the "Prediction" tab, which allows to predict the states of new data, from a modeling performed on another data set.

The interface also has a graphical window, which displays some diagrams, and a window that displays information messages to the user.

Figure 1: Main tab of the application

  1. mapping of the attribute space by representative states 

  2. uHMM : unsupervised Hidden Markov Model 

  3. The standard mode, recommended for non-statisticians, finds and uses the optimal parameters according to the data structure. 

  4. The expert mode allows you to choose the methods and parameters to be used.