Algebraic evaluation of optimization in tumors classification with numerical assessments via a flask-react web...

This study addresses the practical problem of building reliable and interpretable tools to support the early detection of breast and prostate cancers. We investigate how the choice of numerical optimization method affects the training of logistic regression (LR) models for binary cancer classification. In particular, we focus on five widely used algorithms:gradient descent (GD), Newton-Raphson (NR), conjugate gradient (CG), Broyden-Fletcher-Goldfarb-Shanno (BFGS) and limited-memory BFGS (L-BFGS). We further consider ridge (L2) regularization as a standard mechanism to improve numerical stability, control coefficient magnitude, and enhance generalization reliability across datasets. Through algebraic analysis and numerical experiments on two public datasets for breast and prostate cancer, we compare these methods in terms of convergence behavior, computational efficiency, and classification performance. Our results highlight clear trade-offs between runtime, number of iterations, and predictive quality, which are directly relevant when integrating LR into real-world digital health systems. Guided by these insights and by standard practice in the literature, we implement a conventional regularized LR model in a Flask-React web application dedicated to cancer detection. In this platform, the Flask backend manages data processing, model training, and inference, while the React frontend offers an interactive interface for healthcare professionals. Clinicians can enter patient-level features and immediately visualize probabilistic predictions and class labels, illustrating how interpretable statistical models, coupled with appropriate optimization strategies, can be deployed in a web-based decision-support tool for breast and prostate cancer screening.