Early Detection of Breast Cancer: Comparative Analysis of Machine Learning and Deep Learning Algorithms

Breast cancer classification remains a critical challenge in medical diagnostics due to the imbalanced nature of available datasets, where the minority (cancerous malignant) class is often overshadowed by the majority (benign) class. This study proposes a hybrid model based on logistic regression, enhanced with class balancing techniques and ant search optimization, to improve the identification of the malignant class. The model is compared with SVM, Random Forest, and KNearest Neighbors (KNN) across three stages: prediction before diagnosis, at diagnosis and therapy, and post-treatment outcomes. The experiments, conducted on the Jupyter platform using the Wisconsin breast cancer dataset, demonstrate that the hybrid model achieves a high accuracy of 92.98%, significantly reducing false negatives. The study highlights the strengths of logistic regression in providing interpretable results, crucial for clinical decision-making, especially when compared to more complex models like Artificial Neural Networks (ANN). This research offers a reliable and accurate tool for early breast cancer detection and prognosis, contributing to ongoing efforts to enhance patient outcomes through the in