avogogias / MLCut

A visualization support tool for advanced hierarchical clustering analysis. MLCut allows cutting dendrograms at multiple heights/levels. In other words, it allows to set multiple local similarity thresholds in potentially large dendrograms. It uses two coordinated views, one for the dentrogram (radial layout), and another for the original multidimensional data (parallel coordinates). The purpose is to add flexibility and enforce transparency in the process of selecting branches that correspond to the different clusters, while enabling the discovery of visual patterns in the original data.

Instructions:

1. Clone MLCut (if you use a web server save it in a web directory otherwise look at ** below)
2. Edit the "scripts/CSV-TSclust-hclust-rjson.r" file to point to your own CSV data file. Note that the first column that contains the names/ids of your data should be named "ID". Column names of the measurement could be anything you like
3. Execute the R script. You may be asked to install any missing r-packages while running the script. The output will be stored in a JSON file which together with the original CSV file will be used as input to MLCut
4. Edit the "PATH_TO_JSON" and "PATH_TO_CSV" variables in the first lines of "mlcut.js" to match the path and file names of your own .csv and .json data files
5. Access index.html with your web browser **

--

** How to configure Chrome so that it allows XMLHttpRequest (file access from files) in the case you don't run a local web server:

a) Create a Shortcut for Chrome

b) Right Click on Shortcut icon

c) Select Properties

d) Select Shortcut tab

e) Add "--allow-file-access-from-files" flag on Target input e.g. Target: "C:\Program Files (x86)\Google\Chrome\chrome.exe" --allow-file-access-from-files

f) -> Click Apply -> Click OK

g) Open index.html using the Chrome shortcut

--

Find our paper in: https://diglib.eg.org/handle/10.2312/cgvc20161288