Classification of Leukemia Using a Hybrid Approach Based on Temporal Fusion Transformer and XG-Boost
Iranian Journal of Operations Research · 01.08.2025 · 16 (2) · 26-38
DOI: 10.66224/iors.16.2.26 ↗Kurz gefasst
Fachartikel in der Zeitschrift Iranian Journal of Operations Research, online erschienen am 01.08.2025. Verfasst von Sareh Bagheri Matak und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Leukemia is a prevalent and life-threatening cancer, where early detection significantly improves curability.Microarray data, which enables simultaneous measurement of thousands of gene expressions, offers a powerful tool for early diagnosis.However, its high dimensionality and inherent noise complicate analysis, necessitating effective gene selection to enhance accuracy and reduce computational burden.This paper proposes a hybrid two-stage framework integrating feature selection with deep temporal modeling for leukemia subtype classification.First, features are filtered using Mutual Information to retain genes with the strongest statistical association to disease labels.Second, XGBoost performs embedded feature ranking, ensuring stable selection of the most discriminative genes across iterations.Finally, a Temporal Fusion Transformer is employed for classification, efficiently capturing complex temporal patterns within the refined gene set.Evaluated on a real-world microarray dataset comprising 22,284 genes from 640 samples across five leukemia subtypes, the proposed method achieved an accuracy of 99.26%, precision of 99.3%, and recall of 98.9%.A sensitivity analysis, demonstrating the model's stability to parameter variation.The method significantly outperformed baseline models, and state-of-the-art deep learning approaches, while successfully identifying a compact set of biologically relevant genes for differentiation.
Zitieren
APA
Matak, S. B., Askari, E., & Motamed, S. (2025). Classification of Leukemia Using a Hybrid Approach Based on Temporal Fusion Transformer and XG-Boost. Iranian Journal of Operations Research, 16(2), 26-38. https://doi.org/10.66224/iors.16.2.26
Vancouver
Matak SB, Askari E, Motamed S. Classification of Leukemia Using a Hybrid Approach Based on Temporal Fusion Transformer and XG-Boost. Iranian Journal of Operations Research. 2025;16(2):26-38. doi:10.66224/iors.16.2.26
BibTeX
@article{Matak2025,
author = {Matak, Sareh Bagheri and Askari, Elham and Motamed, Sara},
title = {Classification of Leukemia Using a Hybrid Approach Based on Temporal Fusion Transformer and XG-Boost},
journal = {Iranian Journal of Operations Research},
year = {2025},
volume = {16},
number = {2},
pages = {26--38},
doi = {10.66224/iors.16.2.26}
}
EUVIMED