ROBUST MCD BASED DOWN WEIGHTED LASSO REGRESSION FOR BREAST CANCER DIAGNOSIS
JP Journal of Biostatistics · 2026 · 26 (3) · 477-492
DOI: 10.17654/0973514326025 ↗Kurz gefasst
Fachartikel in der Zeitschrift JP Journal of Biostatistics, 2026. Verfasst von K. Manoj und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Breast cancer diagnosis using high dimensional medical data is often affected by biological heterogeneity and the presence of statistical outliers, which can distort model estimation and predictive performance. In this study, we apply Lasso logistic regression combined with a robust preprocessing framework based on the minimum covariance determinant (MCD). Two outlier handling strategies are considered: MCD based hard outlier removal and a robust down weighting approach in which anomalous observations are retained with reduced influence based on robust distances. The results show that both approaches have strong classification performance, whereas the proposed down weighting retains all training observations and improves comparable model stability and classification performance. Overall, the findings indicate that integrating MCD based down weighting with Lasso regularization provides a robust and information preserving practical framework for breast cancer diagnosis and classification using heterogeneous medical data.
Zitieren
APA
Manoj, K., & Tamilarasan, A. (2026). ROBUST MCD BASED DOWN WEIGHTED LASSO REGRESSION FOR BREAST CANCER DIAGNOSIS. JP Journal of Biostatistics, 26(3), 477-492. https://doi.org/10.17654/0973514326025
Vancouver
Manoj K, Tamilarasan A. ROBUST MCD BASED DOWN WEIGHTED LASSO REGRESSION FOR BREAST CANCER DIAGNOSIS. JP Journal of Biostatistics. 2026;26(3):477-492. doi:10.17654/0973514326025
BibTeX
@article{Manoj2026,
author = {Manoj, K. and Tamilarasan, A.},
title = {ROBUST MCD BASED DOWN WEIGHTED LASSO REGRESSION FOR BREAST CANCER DIAGNOSIS},
journal = {JP Journal of Biostatistics},
year = {2026},
volume = {26},
number = {3},
pages = {477--492},
doi = {10.17654/0973514326025}
}
EUVIMED