Subspace Clustering for High-Dimensional Data: A Survey of Methods, Challenges, and Conceptual Frameworks for Future Research

Abstract

Subspace clustering has emerged as a powerful paradigm for analyzing high-dimensional data, where traditional clustering methods struggle due to the curse of dimensionality. By identifying clusters in relevant subsets of dimensions, subspace clustering enhances interpretability, scalability, and robustness in various applications such as bioinformatics, image processing, and IoT. This paper presents a comprehensive survey of existing subspace clustering methods, categorizing them into grid-based, model-based, spectral, and hybrid approaches. We introduce a new taxonomy framework for classifying subspace clustering techniques based on scalability, noise tolerance, and application domains. Additionally, we highlight recent advancements, including deep learning-based subspace clustering, fairness-aware clustering, and real-time streaming data applications. The paper also discusses key challenges such as interpretability, computational complexity, and lack of standardized evaluation metrics, providing insights into future research directions. This survey aims to serve as a roadmap for researchers by consolidating the latest developments and identifying open challenges in subspace clustering.