Data preprocessing using banded structure and image morphology enhancing Boolean matrix factorization

doi:10.1016/j.knosys.2025.114366

Data preprocessing using banded structure and image morphology enhancing Boolean matrix factorization

2025-08-30

0

PRE

AI

DOI：10.1016/j.knosys.2025.114366

Original

Original request for help

Save

Abstract

En 中文

Boolean Matrix Factorization (BMF) is a widely used method for revealing underlying patterns, called factors, in data. In the paper, we propose a novel data preprocessing method that makes patterns more visible, thus simplifying the overall BMF process. The method first reorders the data to reveal a banded structure. Then, it applies image morphology to enhance this structure by emphasizing important information and suppressing less relevant information. We demonstrate the efficacy of our approach through various experimental evaluations, showing that it effectively modifies the data, resulting in fewer, more interpretable factors. The proposed method also allows using more straightforward and faster BMF algorithms while maintaining high-quality results.

Data preprocessing using banded structure and image morphology enhancing Boolean matrix factorization

Abstract

Journal

Organization