A collaboration between Karunia Eka Lestari and Mokhammad Ridwan Yudhanegara from Universitas Singaperbangsa Karawang, Indonesia, Edwin Setiawan Nugraha from President University, Indonesia, and Sisilia Sylviani from Universitas Padjadjaran, Indonesia, has introduced the use of Tucker3 tensor decomposition for analyzing standardized residual hypermatrices in three-way correspondence analysis.
The research problem addressed lies in the complexity of analyzing high-dimensional categorical data, where interactions between multiple variables often remain hidden in traditional two-way analysis methods. This limitation can obscure meaningful patterns in data-driven research across social sciences, education, and applied mathematics.
To address this, the study employs Tucker3 tensor decomposition as a dimensionality reduction and pattern recognition tool. By applying it to standardized residual hypermatrices in three-way correspondence analysis, the approach enhances the ability to detect latent structures, relationships, and variability within complex categorical datasets.
The findings reveal that Tucker3 decomposition significantly improves interpretability of multidimensional correspondence analysis results. It allows researchers to uncover more nuanced associations between categorical variables, supporting advanced data exploration and more accurate conclusions.
This research contributes to multiple United Nations Sustainable Development Goals (SDGs):
- SDG 4 (Quality Education): by offering improved analytical tools for educational and social science data.
- SDG 9 (Industry, Innovation, and Infrastructure): through advancements in mathematical modeling and data science methods.
- SDG 17 (Partnerships for the Goals): by fostering collaboration between Indonesian institutions in mathematics and applied statistics.
Overall, this study demonstrates how tensor-based methods can expand the scope of correspondence analysis, providing powerful tools for high-dimensional data interpretation and research innovation.
20_Mat_2025
