D Santhakumar, Rajaram Gnanajeyaraman, R Elankavi, J Viswanath, I Govindharaj, J Raja
Department of Computer Science and Engineering, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Saveetha University, Tamil Nadu 602105, India.
Department of Computer Science and Engineering, Aarupadai Veedu Institute of Technology, Vinayaka Mission's Research Foundation, Chennai, Tamil Nadu 603104, India.
MethodsX. 2025 Feb 20;14:103239. doi: 10.1016/j.mex.2025.103239. eCollection 2025 Jun.
Gene selection plays a crucial role in the pre-processing of microarray data, aiming to identify a small set of genes that enhances classification accuracy and reduces costs. Traditional methods, such as Genetic Algorithms (GA) and Maximum Relevance Minimum Redundancy (MRMR), have been widely used, but bio-inspired algorithms like Ant Colony Optimization (ACO) and Ant Lion Optimizer (ALO) have shown promising results. These algorithms are based on natural processes: ACO mimics the foraging behavior of ants, while ALO models the hunting strategy of ant-lion larvae. However, both approaches face challenges like premature convergence and inefficient feature space mapping when used individually. To address these issues, this work introduces a hybrid ACO-ALO method, combining the strengths of both algorithms. The proposed hybrid approach enhances feature selection by improving accuracy, reducing computational complexity, and boosting classifier performance. The proposed model, which identifies the optimal feature set for classification using Support Vector Machine (SVM), has achieved an impressive prediction accuracy of 93.94 %. Results on microarray datasets for leukemia prediction show that the hybrid approach outperforms other methods in terms of both effectiveness and efficiency. This work demonstrates the potential of hybrid optimization techniques in bioinformatics for better gene selection and cancer diagnosis.•Hybrid ACO-ALO approach combines strengths of both algorithms for better feature selection.•Enhances classifier performance while reducing computational complexity.•Outperforms traditional methods on leukemia prediction datasets.
基因选择在微阵列数据的预处理中起着至关重要的作用，旨在识别一小部分能够提高分类准确率并降低成本的基因。传统方法，如遗传算法（GA）和最大相关最小冗余（MRMR），已被广泛使用，但蚁群优化（ACO）和蚁狮优化器（ALO）等生物启发算法也显示出了有前景的结果。这些算法基于自然过程：ACO模仿蚂蚁的觅食行为，而ALO模拟蚁狮幼虫的捕猎策略。然而，这两种方法单独使用时都面临过早收敛和特征空间映射效率低下等挑战。为了解决这些问题，这项工作引入了一种ACO-ALO混合方法，结合了两种算法的优势。所提出的混合方法通过提高准确率、降低计算复杂度和提升分类器性能来增强特征选择。所提出的模型使用支持向量机（SVM）识别用于分类的最优特征集，已实现了93.94%的令人印象深刻的预测准确率。白血病预测微阵列数据集的结果表明，混合方法在有效性和效率方面均优于其他方法。这项工作展示了混合优化技术在生物信息学中用于更好的基因选择和癌症诊断的潜力。
•ACO-ALO混合方法结合了两种算法的优势以实现更好的特征选择。
•在降低计算复杂度的同时提升分类器性能。
•在白血病预测数据集上优于传统方法。
