| 摘 要: 特征选择的主要目的是缩减和精炼数据的特征集,使得生成的特征子集可以进一步提高模型的学习精度。针对特征选择这一特定问题,提出了一种二进制登山队优化算法BMTBO(Binary Mountaineering Team-Based Optimization)。该算法属于二进制版本的群智能优化算法,并采用倒“S”形和“V”形数值空间转换函数,实现了在特征选择时,不仅可以降低特征维数,而且可以降低模型学习的误差率。为验证BMTBO算法的实际效果,在15个公共测试数据集上进行实验。实验结果显示,与现有方法相比,BMTBO算法的平均分类准确率最多可提升1百分点,证明所提出的算法在提高模型学习精度方面的可行性与有效性。 |
| 关键词: 二进制群智能优化;登山队优化算法;特征选择;转换函数;数据分类 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 江苏省高校自然科学研究重大项目(22KJA520010);江苏省研究生科研与实践创新计划项目(SJCX23_0273、SJCX24_0318) |
|
| Binary Mountaineering Team-Based Optimization Algorithm and Its Application in Feature Selection |
|
MA Li, GU Lei
|
(School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)
Marry18030101@163.com; njgulei@126.com
|
| Abstract: The primary goal of feature selection is to reduce and refine the feature set of data, allowing the resulting feature subset to improve the learning accuracy of models. In response to this specific issue of feature selection, a Binary Mountaineering Team-Based Optimization (BMTBO) algorithm is proposed. This algorithm is a binary version of a swarm intelligence optimization algorithm and utilizes inverted "S" shape and "V" shaped numerical space transformation functions. This approach not only reduces feature dimensions during feature selection but also reduce models' learning error rate. To validate the practical effectiveness of the BMTBO algorithm, experiments are conducted on 15 public test datasets. The experimental results indicate that, compared to existing methods, the average classification accuracy of the BMTBO algorithm can improve by up to 1 percentage points, demonstrating the feasibility and effectiveness of the proposed algorithm in enhancing model learning accuracy. |
| Keywords: binary swarm intelligence optimization; mountaineering team optimization algorithm; feature selection; transformation function; data classification |
