Author:
Di Nuzzo Cinzia,Ingrassia Salvatore
Abstract
AbstractIn this paper, we present a spectral clustering approach for clustering three-way data. Three-way data concern data characterized by three modes: n units, p variables, and t different occasions. In other words, three-way data contain a t × p observed matrix for each statistical observation. The units generated by simultaneous observation of variables in different contexts are usually structured as three-way data, so each unit is basically represented as a matrix. In order to cluster the n units in K groups, the spectral clustering application to three-way data can be a powerful tool for unsupervised classification. Here, one example on real three-way data have been presented showing that spectral clustering method is a competitive method to cluster this type of data.
Publisher
Springer International Publishing
Reference17 articles.
1. Bocci, L., Vicari, D.: ROOTCLUS: Searching for “ROOT CLUSters” in Three-Way Proximity Data. Psychometrika. 84, 941–985 (2019)
2. Di Nuzzo, C., Ingrassia, S.: A mixture model approach to spectral clustering and application to textual data. Stat. Meth. Appl. Forthcoming (2022)
3. Di Nuzzo, C.: Model selection and mixture approaches in the spectral clustering algorithm. Ph.D. thesis, Economics, Management and Statistics, University of Messina (2021)
4. Di Nuzzo, C., Ingrassia, S.: A joint graphical approach for model selection in the spectral clustering algorithm. Tech. Rep. (2022)
5. Garcia Trillos, N., Hoffman, F., Hosseini, B.: Geometric structure of graph Laplacian embeddings. arXiv preprint arXiv:1901.10651 (2019)