The paper “Comparative Assessment of Triclustering Algorithms for Temporal Data Analysis”, authored by LASIGE’s PhD student Diogo F. Soares and integrated researcher Sara C. Madeira, has been published in Pattern Recognition, a top-ranked journal (Scimago Q1; Google-Scholar rank 6 – Computer Vision & Pattern Recognition). The paper is co-authored by Rui Henriques (INESC-ID and Instituto Superior Técnico – Universidade de Lisboa).

Temporal data analysis involves understanding patterns and trends that change over time, a task that presents unique challenges. Triclustering, which allows for the simultaneous clustering of data along three dimensions—objects, attributes, and time—holds promise in this domain. However, selecting the most appropriate triclustering algorithm for a given dataset can be challenging.

To address this challenge, LASIGE researchers conducted a thorough comparative analysis of state-of-the-art triclustering algorithms. Using synthetic datasets of varying sizes and complexities, they evaluated the performance of these algorithms in uncovering hidden patterns. By considering factors such as dataset characteristics and algorithm capabilities, the study provides valuable insights into the strengths and limitations of each method.

What sets this study apart is its focus on algorithms specifically designed for temporal data analysis. By offering a neutral comparison of these approaches, the research aims to assist practitioners in selecting the most suitable algorithm for their analytical needs.

In a world where data-driven decision-making is increasingly important, understanding temporal dynamics is key. This study contributes to the ongoing exploration of temporal data analysis, offering valuable guidance for researchers and practitioners alike.

The full open-access paper is available here.