Ensemble Based Classification for Classification of Ductal Carcinoma using Reduced Feature Set

Ductal Carcinoma is one of the main causes of death in women. Ductal Carcinoma's affected population has been increasing annually. All living creatures are composed of unique and identifiable cells, which define their physical appearance and purpose. Cells are the basic building block in the human body that construct a particular human organ. In the human body, cells usually replicate themselves in a particular sequence. Bacterial infection in the human body might affect the growth sequence of cells, which enables them to replicate their selves in an uncontrolled sequence. Ductal Carcinoma is a bacterial infection that is caused due to presence of abnormal cells in the duct. It usually occurs due to the outgrowth of cells lining the milk ducts. The chances of ductal carcinoma in humans can be reduced by adopting an early prognosis. The effective prognosis of ductal carcinoma can be done by using learning algorithms such as K-Mean Clustering, K-Nearest Neighbor, Naive Bayes, Neural Network, RBF network, and Support Vector Machine. A number of studies have reported the use and effect of these algorithms However, most of the studies are suffering in terms of accuracy because of insignificant attributes. This study aims to address this concern of accuracy. To execute the intended work, the authors have selected a renowned dataset entitled “Wisconsin Breast Cancer Dataset” which contains all the vital attributes for the effective prognosis of ductal carcinoma such as Bland chromatin, Cell shape/size uniformity, Clump Thickness, Marginal Adhesion, Normal Nuclei, Single Epithelial Cell Size, Bare Nuclei, and Mitosis. The machine learning approach is known to be beneficial in efficiently predicting ductal carcinoma. In this study, we implemented a machine learning approach using the Bagging approach with Naive Bayes. The methodology used has achieved higher accuracy of 97.56 with a lower root mean square error of 0.1612 as compared to the existing study.