The rapid increase in data streams poses significant challenges to their interpretation. Algorithms increasingly target 3D datasets, which plot observations, attributes, and contexts, to capture patterns in the fields of medicine and social networks. One popular approach is that of biclustering, which isolates subsets of data by eliminating one of the dimensions to achieve conceptual simplicity and computational complexity; however, it does so at the cost of completeness and effectiveness. As the authors note, while the more comprehensive approach of triclustering has received scrutiny, it lacks structure. Not only do the authors provide the first comprehensive review of triclustering literature, they go beyond that to contribute to the field by formalizing the triclustering method, comparing it to competing methods, developing a taxonomy and principles for assessing implementations of triclustering, and reviewing practical applications.

Two sections are especially informative and form the article’s core. One is the formulation of the problem that triclustering addresses. The authors effectively present the method using multicolored and 3D illustrations. The other is a comprehensive, concisely written review of triclustering algorithms, including greedy, stochastic, and exhaustive approaches.

The authors complement their theoretical concerns with an analysis of triclustering in real-world areas such as biology, medicine, and social data. This balances out the article’s conceptual concerns.

Researchers in machine learning and in biology, medicine, or social networking will find this article valuable.