Visual Analytics for Temporal Blind Source Separation

This repository contains code for the prototype TBSSvis, a visual analytics app for the Blind Source Separation algorithm gSOBI [1, 1a]. It provides a consistent and intuitive interface to select parameters and explore components.

Please refer to the paper:

N. Piccolotto, M. Bögl, T. Gschwandtner, C. Muehlmann, K. Nordhausen, P. Filzmoser, and S. Miksch: "TBSSvis: Visual Analytics for Temporal Blind Source Separation", Visual Informatics (2022). http://dx.doi.org/10.1016/j.visinf.2022.10.002

Prerequisites: Node.js (>= v12), Docker (>= 19.03.13). Note that the tool itself is called TBSSvis but some things, e.g. environment variables, use TSBSS (observe the additional S).

1. Clone this repository
2. Install frontend dependencies: npm install
3. Build the backend: docker build server --tag tsbss
4. Run the backend with docker: docker run -e TSBSS_DATASET=exrates -d -p 8008:8000 -v $PWD/server/app:/app tsbss - after some preprocessing time the backend runs on port 8008 on your host, check with docker logs if it's working
5. Run the frontend: npx vue-cli-service serve - the frontend runs on port 8080 on your host

TBSSvis should now be accessible at http://localhost:8080/ with the exrates [2] dataset. See next section how to change the dataset.

You can also run the backend server (a plumber API) on your host. Open api.R with RStudio. Install its dependencies and comment out the setwd line. Any dataset name should be hardcoded a few lines below. Open a Terminal in the folder server of this project. Start an R session (just the R command). Then type plumber::pr_run(plumber::pr("api.R", dir="app"), port=8008) and hit return.

Data is loaded from the subfolder server/app/data. Two datasets are included: exrates [2], a dataset of currency exchange rates, and fetal-ecg, a medical dataset. Note that TBSSvis expects datasets to be connected to a Gregorian calendar with resolution up to milliseconds.

To use a custom dataset, place a CSV file in the folder mentioned above. The first column must contain the date and be named date. The column names should be preferably short, we developed TBSSvis with 3 characters length.

Then, when starting the backend, point TBSSvis to your dataset with the TSBSS_DATASET environment variable. E.g., if your data is in the file server/app/data/customdata.csv, use -e TSBSS_DATASET=customdata in the docker command in step 4 of the previous section.

• [1] J. Miettinen, M. Matilainen, K. Nordhausen, and S. Taskinen, “Extracting Conditionally Heteroskedastic Components using Independent Component Analysis,” Journal of Time Series Analysis, vol. 41, no. 2, pp. 293–311, 2020, doi: 10.1111/jtsa.12505.
• [1a] https://cran.r-project.org/web/packages/tsBSS/index.html
• [2] https://cran.r-project.org/web/packages/stochvol/stochvol.pdf, see section exrates