改进二进制沙丘猫群优化特征选择算法

Improved Binary Sand Cat Swarm Optimization Feature Selection Algorithm

摘要: 特征选择在机器学习的分类任务中被广泛应用，选择出的特征子集会直接影响后续学习算法的性能。针对沙丘猫群优化算法(SCSO)全局搜索能力弱、收敛速度慢问题，本文提出一种改进的二进制沙丘猫群优化特征选择算法。首先改进控制沙丘猫在搜索阶段和攻击阶段转换参数的调整方法，使用两阶段的改进收敛因子策略代替线性递减策略，以提升算法的全局搜索能力。其次受PSO算法位置更新公式的启发，引入社会学习因子和认知学习因子策略，提高算法的收敛速度。为了验证新提出算法在求解特征选择问题上的性能，本文选择了4种经典算法在8个UCI数据集上进行了对比测试，实验结果表明新提出算法的性能优于对比算法。

Abstract: Feature selection is widely used in classification tasks of machine learning, and the selected feature sets directly affect the performance of subsequent learning algorithms. Aiming at the issues of weak global search ability and slow convergence speed of Sand Cat Swarm Optimization (SCSO), an improved binary sand cat swarm optimization feature selection algorithm is proposed in this paper. Firstly, the adjustment method of controlling the transition parameters of sand cat in the search phase and attack phase is improved.This method employs a two-stage improved convergence factor strategy, replacingthe linear decrement strategy, aiming to enhance the algorithm’s global search capability. Secondly, inspired by the position update formula of the PSO algorithm, social learning factor and cognitive learning factor strategies are introduced to improve the convergence speed of the algorithm. In order to verify the performance of the newly proposed algorithm in solving the feature selection problem, this study conducted comparative tests on eight UCI datasets using four classical algorithms. The experimental results demonstrate that the performance of the newly proposed algorithm outperforms the compared algorithms.