Text Clustering and Feature Selection Algorithms

Text clustering is an unsupervised learning technique that organises documents into coherent groups based on shared linguistic and semantic characteristics. The effectiveness of clustering depends critically on the representation of texts, which often involves high-dimensional vectors of terms or features. Uncurbed dimensionality can introduce noise and redundancy, undermining cluster quality and computational efficiency. Feature selection addresses this challenge by identifying a subset of informative words or attributes, thereby enhancing the discriminative power of clustering algorithms. Traditional clustering methods such as k-means, hierarchical clustering and density-based approaches have been complemented by optimisation strategies to overcome sensitivity to initialisation and parameter choices. Contemporary research emphasises hybrid frameworks in which meta-heuristic search techniques simultaneously refine feature subsets and cluster assignments, yielding robust solutions across large and heterogeneous corpora. These advances bear global significance, underpinning applications in search engines, social media analytics and biomedical text mining, where accurate organisation of vast textual resources is paramount.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent studies have explored novel optimisation frameworks to enhance both clustering performance and dimensionality reduction. A 2024 investigation deployed a wind-driven optimisation algorithm as a dual-purpose tool for multi-objective feature selection and clustering. By eliminating over half of the original features, the method achieved a marked improvement in cluster coherence and F-measure values compared with standard meta-heuristic baselines. A 2021 survey of optimisation algorithms in big-data text clustering reviewed genetic algorithms, particle swarm optimisation, harmony search and their hybrid variants. It identified best practices for parameter setting, discussed scalability challenges and proposed future directions for unified frameworks that balance exploration and exploitation. In another 2021 comparative study, several swarm intelligence methods—including particle swarm optimisation, grey wolf optimisation and bat algorithms—were benchmarked against k-means on diverse document collections. The swarm-based approaches consistently outperformed traditional clustering in terms of cluster purity and stability, underscoring the practical benefits of decentralised search strategies in managing the complexity of real-world text corpora.

Text Clustering and Feature Selection Algorithms publication trend

The graph below shows the total number of articles in text clustering and feature selection algorithms across all publications each year (not limited to Nature Index journals).

Technical terms

Text clustering: Unsupervised partitioning of documents into groups according to content similarity.

Feature selection: Process of identifying and retaining the most informative variables or terms for analysis.

Dimensionality reduction: Techniques that transform high-dimensional data into a lower-dimensional representation while preserving key structures.

Meta-heuristic algorithm: High-level, problem-independent strategies that guide subordinate heuristics to explore complex search spaces efficiently.

Swarm intelligence: Collective behaviour of decentralised, self-organised agents—often nature-inspired—applied to optimisation problems.

References

1. Multi-objective of wind-driven optimization as feature selection and clustering to enhance text clustering. International Journal of Data and Network Science (2024).
2. Advances in Meta-Heuristic Optimization Algorithms in Big Data Text Clustering. Electronics (2021).
3. Swarm Intelligence Algorithms in Text Document Clustering with Various Benchmarks. Sensors (2021).