基于改进徒步优化算法的变精度邻域粗糙集特征选择研究

Research on Variable Precision Neighborhood Rough Set Feature Selection Based on Improved Hiking Optimization Algorithm

摘要: 特征选择是数据降维的有效方法，粗糙集理论是处理不确定性信息的有效工具，在粗糙集理论中，变精度邻域粗糙集是处理连续型数值信息系统的有效工具，在分类任务中，变精度邻域粗糙集可以为属性集提供重要性度量。徒步优化算法是群智能优化算法中的一种，可以寻找组合最优解，本文针对徒步优化算法无法动态平衡局部与全局寻优能力，容易陷入局部最优的问题，提出改进徒步优化算法，并提出基于改进徒步优化算法的特征选择算法。为了验证所提算法的有效性，在8个UCI数据集上进行了基于改进徒步优化算法的特征选择算法的对比实验，通过实验验证了所提出算法的有效性和可行性。

Abstract: Feature selection is an effective method for data dimensionality reduction, and rough set theory is an effective tool for dealing with uncertainty information. In rough set theory, variable precision neighborhood rough set is an effective tool for dealing with continuous numerical information systems. Hiking optimization algorithm is one of the swarm intelligence optimization algorithms, which can find the combined optimal solution. In this paper, aiming at the problem that the hiking optimization algorithm cannot dynamically balance the local and global optimization ability, and is easy to fall into the local optimum, an improved hiking optimization algorithm is proposed. A feature selection algorithm based on improved hiking optimization algorithm was proposed. In order to verify the effectiveness of the proposed algorithm, comparative experiments of feature selection algorithm based on improved hiking optimization algorithm are carried out on 8 UCI data sets, and the effectiveness and feasibility of the proposed algorithm are verified by experiments.