Metaheuristic Optimization Techniques for Feature Selection and Data Clustering

Metaheuristic optimisation techniques, inspired by natural phenomena and collective intelligence, have emerged as pivotal tools for reducing dimensionality and uncovering intrinsic structure within complex datasets. In feature selection, these methods aim to identify the most informative attributes by navigating vast combinatorial spaces without exhaustive search, thereby improving classifier accuracy and reducing computational burden. In data clustering, metaheuristics partition observations into cohesive groups by iteratively refining candidate clusterings to maximise intra‐cluster similarity and inter‐cluster separation. Key strategies involve balancing exploration of novel regions of the search space with exploitation of promising solutions, often achieved through adaptive operators or hybridised schemes. Recent advances have focused on binary encodings and transfer functions to handle discrete selection tasks, multi‐trial vector approaches to prevent premature convergence, and domain‐specific enhancements, yielding robust performance across medical diagnostics, social network analysis and engineering design.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent developments in feature selection have seen the introduction of a mutation‐augmented Black Hole Optimisation (MBHO) approach that integrates inversion mutation and enhanced objective functions accounting for feature‐label correlation. Evaluated across fourteen benchmark datasets, this method delivers superior classifier performance by reducing redundancy and preserving relevant features while maintaining a dynamic balance between global exploration and local exploitation. Another influential work proposed binary variants of the Aquila Optimiser (SBAO and VBAO) tailored for medical datasets, including a COVID-19 case study. These algorithms employ S- and V-shaped transfer functions to convert continuous search trajectories into binary feature subsets, achieving high classification accuracy with minimal features. In the clustering domain, a discrete Moth‐Flame Optimisation algorithm for community detection (DMFO-CD) adapts continuous solution representations via locus‐based adjacency encoding. It incorporates customised crossover and mutation operators to optimise modularity, demonstrating competitive performance on real‐world networks by accurately identifying communities without prior knowledge of their number.

Metaheuristic Optimization Techniques for Feature Selection and Data Clustering publication trend

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

Technical terms

Metaheuristic optimisation: A family of stochastic search methods inspired by natural processes, used to find near-optimal solutions in complex search spaces.

Feature selection: The process of selecting a subset of relevant variables from a larger set to improve model performance and reduce overfitting.

Data clustering: The task of grouping data points into clusters such that items within each cluster share higher similarity than those in different clusters.

Exploration and exploitation: Complementary search behaviours where exploration seeks diverse solutions and exploitation focuses on refining the best candidates.

Transfer function: A mapping mechanism in binary metaheuristics that converts continuous algorithmic updates into discrete selection decisions.

References

1. Investigating the Performance of a Novel Modified Binary Black Hole Optimization Algorithm for Enhancing Feature Selection. Applied Sciences (2024).
2. Binary Aquila Optimizer for Selecting Effective Features from Medical Data: A COVID-19 Case Study. Mathematics (2022).
3. DMFO-CD: A Discrete Moth-Flame Optimization Algorithm for Community Detection. Algorithms (2021).
4. MTV-MFO: Multi-Trial Vector-Based Moth-Flame Optimization Algorithm. Symmetry (2021).