Optimized feature subset selection and classification of breast cancer with mammogram image using machine learning technique

Abstract

Breast cancer is a serious disease for women due to high death rate. Different types of tumours formed in the breast tissues and generally, screening process is used to identify the type of tumour. Most of the breast cancer initiates in the ducts or lobules of the breast. In the categorization of breast cancer dataset, reliability is the primary objective to identify breast cancer. All breast abscesses are not malignant, and the entire benign abscess does not progress to cancer. The breast cancer can be diagnosed accurately using Machine Learning (ML) technology in consideration with optimal attributes with multidimensional classification. The most essential feature subset selection in pre-processing step with immaterial and incomplete features is a concern to handle high dimensional data. Also design of a classifier should be effective for huge dataset with respect to the time complexity. One of the important data sets used for breast cancer detection is the MAMMOSET data set. Mammography images look like masses or calcified for malignant tumours. Benign tumours are visualized as rounded and smooth. Calcified masses are characterized through various shapes, thickness, granular, popcorn, or ring, and their density is greater and parturition is widespread. In later stages, the malignant tumours contain needle-like shape in mass, and edges are uneven and normally blurred. The geomorphology calcification is typically less sandy, linear or branched, using dissimilar shapes and sizes; distribution is frequently solid or gathered under linear fashion. Based on low dissimilarity in the mammogram images, it is hard for clinicians to create a right diagnosis. Also, the diagnostic performance of physicians depends on the mammogram images. Generally, in the existing works, computer processors related to large memory and rapid buses for most of the medical imaging applications which is practically tough.Machine learning algorithms are utilized to remove the effects of multidimensional feature characteristics of images. This