BREAST CANCER DETECTION USING LOGISTIC REGRESSION: A COMPARATIVE STUDY OF OPTIMIZATION AND FINE-TUNING METHODS

Abstract

Breast cancer is one of the most prevalent and life-threatening diseases worldwide, making accurate early diagnosis critical for improving patient outcomes. This study investigates the effectiveness of logistic regression models for breast cancer detection using the Wisconsin Breast Cancer (Diagnostic) Dataset. Various optimization techniques, including Ordinary Least Squares, Gradient Descent, Newton’s Method, L2 Regularization, Stochastic Gradient Descent (SGD), and GridSearch-based fine-tuning, were applied to determine their impact on model performance. Standard evaluation metrics—Accuracy, Precision, Recall, and F1-score—were used to compare methods. The results demonstrate that GridSearch-based fine-tuning consistently yields the highest overall performance, highlighting the importance of systematic hyperparameter optimization in enhancing model robustness and predictive accuracy. These findings emphasize that even classical models like logistic regression can achieve state-of-the-art results in medical data analysis when combined with appropriate optimization strategies.