An Ensemble Machine Learning based Framework for Early Detection of Breast Cancer

Abstract

Breast cancer is a disease characterized by the abnormal growth of breast cells, with different types based on the origin and spread of malignant cells. The most common types include infiltrative ductal carcinoma, which starts in the breast ducts and spreads to surrounding tissues, and infiltrative lobular carcinoma, which originates in the lobules and can metastasize to other parts of the body. Given the increasing interest in artificial intelligence for medical diagnostics, various machine learning techniques have been employed to predict breast cancer. In this study, the Wisconsin Breast Cancer Dataset (WBCD) from the UCI Machine Learning Repository was utilized, containing 30 extracted features, including mean, standard error, and worst values for various attributes. To evaluate model performance, key metrics such as specificity, F1-score, sensitivity, and accuracy were analyzed. A stacked ensemble classifier was developed using Decision Tree, AdaBoost, Gaussian NB, and MLP classifiers, achieving a high accuracy of 96.66%, surpassing existing approaches. The results indicate that the proposed ensemble model effectively distinguishes between malignant and benign cancer cells, facilitating early detection and improving treatment outcomes. Additionally, this ensemble approach can be adapted to other medical and classification problems, demonstrating its broader applicability