TREENSIGHT

Treensight is a generic and extensible graphical user interface that allows interactive visualization and manipulation of clustering and decision trees with keeping a tight link with the data. Treensight ensures an efficient collaboration user/machine by offering a better control on the analyzed data and a detailed insight into each cluster.

Treensight is a multiplatform application (macOS, Linux, Windows). It has three components:

• The Core (Javascript): it proposes visualization, filtering and exploration technique up to 1M data.

• An extensible library of scripts (Python): implementation of Machine Learning, analysis and manipulation functions.

• Communication protocole: ensures the link between the Core and the library of scripts.

Treensight has four main features:

• Dynamic content: information about the selected cluster is displayed in the right-hand panel: name, statistics about the instances, distances, etc.

• Instances repartition: the left-hand panel allows the creation of « View points ». In each one the user selects coloring rules and visualization modes that will be displayed in the central panel.

• Variables comparison: user can compare the statistics of a variable coming from different clusters.

• Clusters comparison: in the same fashion, user can compare the clusters and their evolution.

A first cross-platform (Windows, Linus, MacOS) version of the software is released, with restricted access. This software is used in several projects by researchers from various disciplines.

• Science and technology: classification, research of indicators, maintenance...
• Medical and health field: cohort analysis, sensitivity/specificity study...
• Marketing and economy: impact of a campaign, user profiling...
• Sociology: study of populations and their evolution, analysis of surveys...