Related Experiment Videos
Two-mode clustering methods: a structured overview
Iven Van Mechelen1, Hans-Hermann Bock, Paul De Boeck
1Psychology Department, University of Leuven, Tiensestraat 102, B-3000 Leuven, Belgium. iven.vanmechelen@psy.kuleuven.ac.be
Statistical Methods in Medical Research
|November 2, 2004
Summary
This study overviews two-mode clustering methods for simultaneously grouping rows and columns in data matrices. It structures these methods by cluster characteristics and model types, using psychiatric symptom data as an example.
Area of Science:
- Data Mining
- Machine Learning
- Statistical Analysis
Background:
- Simultaneous clustering of rows and columns (two-mode clustering) is crucial for analyzing rectangular data matrices.
- Existing methods lack a unified structure, hindering comparative analysis and application.
- Understanding the interplay between row, column, and data clusters is essential.
Purpose of the Study:
- To provide a structured overview of existing two-mode clustering methods.
- To categorize these methods based on key structuring principles.
- To demonstrate the application of these methods using real-world data.
Main Methods:
- Categorization of two-mode clustering techniques based on cluster nature (row, column, data) and model/loss function.
- Systematic review and synthesis of relevant literature.
- Application of selected methods to psychiatric symptom data.
Main Results:
- A clear taxonomy of two-mode clustering approaches is presented.
- Methods are differentiated by their handling of cluster types and underlying models.
- Illustrative analyses highlight the practical utility in psychiatric research.
Conclusions:
- The structured overview facilitates method selection and development in two-mode clustering.
- Two-mode clustering offers valuable insights into complex datasets, as shown in the psychiatric patient example.
- Further research can build upon this framework for advanced data analysis.