Machine Learning Techniques for Breast Cancer Prediction in Imbalanced Datasets: A Systematic Review

Abstract

Background: Class imbalance remains a central methodological obstacle in machine learning (ML)-based breast cancer prediction, where the clinically critical minority class is systematically underrepresented relative to the majority class. A systematic review by Ghavidel and Pazos synthesized ML approaches to this problem across breast cancer prediction literature published between 2008 and 2023. Given the continued rapid publication of ML and deep-learning methods for breast cancer classification, detection, and prognosis since that search closed, an updated synthesis is warranted. Objective: This manuscript reports the protocol, conceptual framework, and methodology for a systematic review that updates and extends the prior evidence synthesis, focusing on studies published between September 2023 and August 2026 that apply ML techniques to breast cancer prediction under conditions of class imbalance. Methods: The review follows PRISMA 2020 reporting guidance. A multi-database search strategy, eligibility criteria, and a standardized data-extraction framework, harmonized where possible with the prior review, are defined and reported transparently, together with an expanded conceptual background covering the taxonomy of imbalance-handling strategies and the evaluation metrics appropriate to imbalanced classification. Screening and extraction of individual studies, and the resulting comparative synthesis, dataset and algorithm characterization, quality assessment, and discussion, will be completed and reported in subsequent parts of this manuscript once the documented multi-database search has been executed and verified; no screening, inclusion, or performance figures are reported until they can be confirmed against source records. Conclusion (protocol stage): This manuscript establishes a transparent, reproducible, non-fabricated methodological and conceptual foundation for evaluating how the recent literature has addressed class imbalance in breast cancer prediction.