Enhancing Feature Selection with PSO-RK-Gaussian Algorithm for Cancer Prediction

The PSO-RK-Gaussian algorithm enhances traditional Particle Swarm Optimization (PSO) by integrating Runge-Kutta (RK) numerical methods and Gaussian sampling techniques, specifically to improve feature selection in complex biomedical datasets. PSO uses a population of potential solutions called particles, with each particle representing a candidate solution within the search space. These particles iteratively adjust their positions based on two main factors: their own best-known positions (local best) and the best-known positions of the entire swarm (global best). This dual adjustment allows the particles to navigate toward promising regions within the search space, effectively balancing exploration and exploitation to find optimal solutions. Our enhanced method demonstrates clear superiority over traditional PSO and other baseline techniques in both feature selection quality and predictive modelling accuracy. Through rigorous experimentation, the PSO-RK-Gaussian algorithm shows significant improvements in key evaluation metrics, including accuracy, precision, recall, F1-score, and AUC-ROC. These results underscore the algorithm's potential for applications in bioinformatics, particularly for analysing large and complex biomedical datasets. Additionally, our method provides a robust framework that can be adapted to various optimization problems beyond bioinformatics. This adaptability and the demonstrated improvements in predictive modelling suggest a broad applicability of the PSO-RK-Gaussian algorithm across different domains requiring efficient and effective feature selection.