Theoretical (Analytical):

Practical (Implementation):

Literature Work:

---

Overview

The project will develop a client-server architecture for clustering computations (backend) with a web-based visual interface for clustering workflows. It is inspired by an existing approach for cluster analysis using self-organizing maps.

Problem Statement

Iterative refinement and filter steps are needed to successively approach a desired visualization of a certain dataset. However, a variety of different clustering algorithms exists. This project aims to embed different algorithms into a common clustering workflow approach that allows the user to iteratively switch and compare different results.

Tasks

• Get familiar with the state of the art data mining and machine learning algorithms and the current system. 
• Design and implement several interactive algorithm visualizations and embed them in a workflow.
• Extend the backend (java) with a variety of algorithms and develop a web-based visual interface (html/javascript/d3). 

Requirements

• Basic knowledge about machine learning, data mining algorithms and visual analytics
• Advanced programming skills in Java
• Web programming skills (HTML/Javascript/d3.js)
• Good conceptual skills (software architectures)
• Basic knowledge in REST, Client-Server communication, Ajax/Jersey, JSON
• Useful: Git, Maven

Scope/Duration/Start

• Scope: Master
• 6 Month Project
• 6 Month Thesis

Contact

• Matthias Kraus

References

• D. Sacha and M. Kraus and J. Bernard and M. Behrisch and T. Schreck and Y. Asano and D. A. Keim:
  SOMFlow: Guided Exploratory Cluster Analysis with Self-Organizing Maps and Analytic Provenance. IEEE Transactions on Visualization and Computer Graphics (2017)
• Dominik Sacha, Michael Sedlmair, Leishi Zhang, John Aldo Lee, Jaakko Peltonen, Daniel Weiskopf, Stephen C. North, Daniel A. Keim:
  What you see is what you can change: Human-centered machine learning by interactive visualization. Neurocomputing 268: 164-175 (2017)