Evaluating Support Vector Machine and Logistic Regression: A Machine Learning-Based Statistical Study for Breast Cancer Diagnosis

In the context of binary classification for breast cancer diagnosis, this paper offers a comparative statistical analysis of two popular classification techniques, Support Vector Machine (SVM) and Logistic Regression. In the case of breast cancer treatment, where missed diagnoses (false negatives) can have serious repercussions, early and accurate diagnosis is essential to improve patient outcomes. By seeing intricate patterns in medical data that may be difficult for conventional statistical techniques to detect, machine learning models can improve diagnostic accuracy. Examining the distinctions, advantages, and disadvantages of the two statistical models for identifying tumor samples as benign or malignant is the goal of this study. SVM is statistically examined for its ability to optimize the margin between classes, which can enhance the robustness of classification, especially in high- dimensional datasets. Its adaptability in managing non-linear data structures using different kernel functions is also evaluated. Conversely, Logistic Regression is favored for issues where linear separability is assumed due to its interpretability, computational efficiency, and dependence on a linear connection between characteristics and outcomes. The models are compared on a breast cancer dataset using statistical measures like accuracy, precision, recall, F1-score, and confidence intervals. Although both models perform well, SVM exhibits marginally improved accuracy and precision, rendering it more appropriate for minimizing false positives. To determine whether the accuracy differences between SVM and Logistic Regression are statistically significant, a paired t-test is also used. The assertion that both algorithms are feasible choices for breast cancer diagnosis depending on the dataset utilized is further supported by the paired t-test results, which show no statistically significant difference between the two models. This study offers evidence-based insights to choose the best classification model for breast cancer detection based on both statistical performance and clinical relevance by combining statistical significance testing with performance indicators.