基于自适应森林优化算法的特征选择算法  

Feature selection algorithm based on adaptive forest optimization algorithm

作　　者：黄君策 石林[1] 顾玉宛[1] 李宁[1] 庄丽华[1] 徐守坤[1] HUANG Jun-ce;SHI Lin;GU Yu-wan;LI Ning;ZHUANG Li-hua;XU Shou-kun(School of Computer Science and Artificial Intelligence&Aliyun School of Big Data&School of Software,Changzhou University,Changzhou 213164,China)

机构地区：[1]常州大学计算机与人工智能学院、阿里云大数据学院软件学院,江苏常州213164

出　　处：《计算机工程与设计》2023年第2期425-431,共7页Computer Engineering and Design

基　　金：江苏省石油化工过程关键设备数字孪生技术工程研究中心基金项目(苏发改高技发[2019]1125号);2018年高校“青蓝工程”优秀青年骨干教师培养基金项目(苏教师〔2018〕12号)。

摘　　要：森林优化特征选择算法(FSFOA)表现出色,但初始化方法的盲目性以及更新机制的局限性限制了该算法的性能。对FSFOA不足之处加以改进,提出自适应森林优化特征选择算法(AFSFOA)。在初始化过程中,加入特征权重评估算法;在更新机制上,使用自适应参数选择策略以及贪心搜索策略替代原始的更新机制。在不同维度的数据集上进行实验,对比实验结果表明,与FSFOA算法以及近年来提出的较高效的特征选择算法进行对比,在准确率以及维度约简上,AFSFOA算法有很强的竞争力。The forest optimization feature selection algorithm(FSFOA) performs well, which is still limited by the blindness of the initialization method and the limitation of the update mechanism. The adaptive feature selection forest optimization algorithm(AFSFOA) was proposed to improve the deficiencies of FSFOA. Feature weight evaluation algorithm was added in the initialization. Adaptive parameter selection strategy and greedy search strategy were used to replace the original update mechanism in the update mechanism. Experiments were conducted on various-dimensional datasets. The comparison results show that the AFSFOA algorithm is competitive in terms of accuracy and dimensional simplification, compared with the FSFOA algorithm and more efficient feature selection algorithms proposed in recent years.

关 键 词：数据挖掘 特征选择 初始化策略 特征权重评估算法 更新机制 贪心算法 森林优化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]