Integrated Ensemble Strategy for Breast Cancer Detection Using Dimensionality Reduction Technique

Đã lưu trong:

Chi tiết về thư mục
Xuất bản năm:	ADCAIJ : Advances in Distributed Computing and Artificial Intelligence Journal vol. 14 (2025), p. e31899-e31916
Tác giả chính:	Zulfikar Ali Ansari
Tác giả khác:	Arif, Mohammad, Nagendra Babu Rajaboina, Shaikh, Anwar Ahamed, Singh, Yaduvir
Được phát hành:	Ediciones Universidad de Salamanca
Những chủ đề:	Datasets Singular value decomposition Principal components analysis Machine learning Support vector machines Multilayers Public health Multilayer perceptrons Breast cancer Decision trees
Truy cập trực tuyến:	Citation/Abstract Full Text - PDF
Các nhãn:	Thêm thẻ Không có thẻ, Là người đầu tiên thẻ bản ghi này!

Miêu tả
Bài tóm tắt:	Breast cancer remains a critical global health concern, requiring advanced and reliable diagnostic methods for early detection and effective intervention. This work introduces an integrated ensemble framework that combines multiple dimensionality reduction (DR) techniques, including Principal Component Analysis (PCA), Non-negative Matrix Factorization (NMF), and Singular Value Decomposition (SVD), with robust machine learning (ML) classifiers for improved breast cancer detection. The publicly available Wisconsin Breast Cancer Dataset (WBCD) was utilized, with rigorous data preprocessing performed to address missing values, anomalies, and class imbalance through stratified sampling and median imputation. To mitigate overfitting and underfitting, dimensionality reduction was coupled with cross-validation and ensemble strategies. The predictive performance of Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Multi-Layer Perceptron (MLP) was systematically evaluated. Experimental results show that SVM consistently achieves a maximum accuracy of 97. 9 % across all applied DR techniques, while MLP and LR also reach 97. 9 % accuracy with PCA and NMF, though MLP exhibits performance variability depending on the selected DR method. The findings provide practical guidance for healthcare practitioners and researchers, supporting the adoption of explainable and scalable AI-driven diagnostic tools. Limitations include the reliance on a single dataset and the need for further validation on larger and more diverse clinical cohorts. Future work will focus on enhancing model interpretability, external validation, and real-world deployment in resource-constrained settings.
số ISSN:	2255-2863
DOI:	10.14201/adcaij.31899
Nguồn:	Advanced Technologies & Aerospace Database