A systematic review of machine learning algorithms for breast cancer detection

Abstract

Breast cancer is one of the leading causes of death and morbidity among women worldwide. Identifying cancerous cells remains a complex and time-consuming task, particularly when performed manually by radiologists or pathologists, contributing to high diagnostic costs. The absence of a reliable, standardized predictive model often hinders timely and accurate diagnosis. This systematic review explores various machine learning approaches — including eXtreme Gradient Boosting (XGBoost), Naïve Bayes, Support Vector Machine (SVM), Logistic Regression, Decision Tree, and k-Nearest Neighbors (KNN) — for classifying breast tumors as malignant or benign. It synthesizes findings from existing literature, comparing model performance based on key evaluation metrics such as accuracy, precision, recall, and F1-score. Multiple reviewed studies report that machine learning models can achieve high diagnostic accuracy. These models may improve diagnostic confidence and accelerate result interpretation. This review also highlights common limitations, such as dataset availability, class imbalance, model interpretability, and generalizability across diverse populations. The paper concludes by outlining future directions to enhance the clinical applicability, trustworthiness, and integration of ML-based diagnostic systems.