Aswal Shobha, Ahuja Neelu Jyothi, Mehra Ritika
Department of Computer Science and Engineering, VM Singh Bhandari Uttarakhand Technical University, Dehradun, India.
Department of Systemics, School of Computer Science, University of Petroleum and Energy Studies, Dehradun, India.
Sci Rep. 2025 May 29;15(1):18954. doi: 10.1038/s41598-025-02954-4.
Due to imbalanced data values and high-dimensional features of lung cancer from CT scans images creates significant challenges in clinical research. The improper classification of these images leads towards higher complexity in classification process. These critical issues compromise the extraction of biomedical traits and also design incomplete classification of lung cancer. As the conventional approaches are partially successful in dealing with the complex nature of high-dimensional and imbalanced biomedical data for lung cancer classification. Thus, there is a crucial need to develop a robust classification technique which can address these major concerns in the classification of lung cancer images. In this paper, we propose a novel structural formation of the oblique decision tree (OBT) using a swarm intelligence technique, namely, the Binary Bat Swarm Algorithm (BBSA). The application of BBSA enables a competitive recognition rate to make structural reforms while building OBT. Such integration improves the ability of the machine learning swarm classifier (MLSC) to handle high-dimensional features and imbalanced biomedical datasets. The adaptive feature selection using BBSA allows for the exploration and selection of relevant features required for classification from ODT. The ODT classifier introduces flexibility in decision boundaries, which enables it to capture complex linkages between biomedical data. The proposed MLSC model effectively handles high-dimensional, imbalanced lung cancer datasets using TCGA_LUSC_2016 and TCGA_LUAD_2016 modalities, achieving superior precision, recall, F-measure, and execution efficiency. The experiments are conducted in Python to evaluate the performance metrics that consistently demonstrate enhanced classification accuracy and reduced misclassification rates compared to existing methods. The MLSC is assessed in terms of both qualitative and quantitative measurements to study the capability of the proposed MLSC in classifying the instances more effectively than the conventional state-of-the-art methods.
由于CT扫描图像中肺癌数据值的不平衡和高维特征,给临床研究带来了重大挑战。这些图像的不当分类导致分类过程更加复杂。这些关键问题影响了生物医学特征的提取,也导致肺癌分类设计不完整。由于传统方法在处理用于肺癌分类的高维和不平衡生物医学数据的复杂性质方面仅取得了部分成功。因此,迫切需要开发一种强大的分类技术,以解决肺癌图像分类中的这些主要问题。在本文中,我们提出了一种使用群体智能技术,即二进制蝙蝠群算法(BBSA)的斜决策树(OBT)的新颖结构形式。BBSA的应用在构建OBT时能够以具有竞争力的识别率进行结构改革。这种集成提高了机器学习群体分类器(MLSC)处理高维特征和不平衡生物医学数据集的能力。使用BBSA的自适应特征选择允许从ODT中探索和选择分类所需的相关特征。ODT分类器在决策边界中引入了灵活性,使其能够捕捉生物医学数据之间的复杂联系。所提出的MLSC模型使用TCGA_LUSC_2016和TCGA_LUAD_2016模式有效地处理高维、不平衡的肺癌数据集,实现了卓越的精度、召回率、F值和执行效率。在Python中进行实验以评估性能指标,与现有方法相比,这些指标始终显示出更高的分类准确率和更低的误分类率。从定性和定量测量两方面对MLSC进行评估,以研究所提出的MLSC在比传统的最先进方法更有效地对实例进行分类方面的能力。