Data-Driven Breast Cancer Diagnosis: a Comparative Study

Breast cancer, a widespread and deadly disease, requires effective and swift diagnostic methods for early detection and treatment. This study utilizes Data analysis methodologies diagnose mammary tumors as either malignant or benign, with the goal of improving survival rates through timely identification. Patient cases were analyzed, unveiling relationships between the characteristics of the condition. Various classification algorithms were compared using diverse data extracted through matching input features from a shared dataset, aiming to determine the most accurate method for predicting malignant or benign tumors. The research introduced an innovative approach to patient diagnosis through an Internet-based service, offering an alternative to traditional doctor-dependent diagnoses. Additionally, the study comprehensively analyzes various binary classification algorithms, including Neural Networks (NN), Random Forest (RF), Support Vector Machine (SVM), and Naive Bayes (NB). The dataset, encompassing 569 samples from Wisconsin, comprises 30 distinct features for diagnosing breast tumors. The study followed a supervised approach, dividing the dataset into a training group of 455 samples for model development and 144 test samples were used to evaluate the proposed model. A set of performance indicators were applied to evaluate the achievement of each classification method, revealing an effectiveness of the selected techniques. Notably, the Support Vector Machine method achieved exceptional performance with a 100% efficiency rating.