.png){kind=icon}
.png){kind=icon}
Return
RLDOS: a novel binary metaheuristic framework with reinforcement learning-driven operator selection for effective feature selection in high-dimensional biomedical data
.png){kind=icon}
PRE
AI
B
V
B
S
DOI：10.1007/s10586-026-06417-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
High-dimensional biomedical data—particularly in cancer diagnosis—poses significant challenges to machine learning models due to the curse of dimensionality and the presence of redundant or irrelevant features. To address this issue, we propose RLDOS (Reinforcement Learning-Driven Operator Selection), a novel binary metaheuristic framework for feature selection. RLDOS integrates a reinforcement learning mechanism modeled as a multi-armed bandit (MAB) problem to guide the dynamic selection of variation operators. Through an adaptive ε-greedy strategy and a reward memory buffer, RLDOS continually learns which operators are most effective in improving solution quality. The framework also incorporates adaptive mutation control, elitism, and stagnation-aware diversification to balance exploration and exploitation. Experiments conducted on fourteen benchmark cancer datasets show that RLDOS achieves competitive and often improved performance compared with several well-known binary optimizers in terms of classification accuracy, number of selected features, and fitness value. Statistical tests, including the Friedman and Wilcoxon signed-rank tests, further support the stability and effectiveness of the proposed approach. These results indicate that RLDOS is a promising framework for binary feature selection in high-dimensional biomedical data analysis.
Keywords:
RLDOS (Reinforcement Learning-Driven Operator Selection)
Binary optimization
Metaheuristics
Feature selection
High-dimensional biomedical data
Journal
C
IF:
4.1
Papers:
4.8K
Citations:
7.5K
Organization
Cited Papers