Publication

Abstract : Due to the prevailing lifestyle, reduced interest in mother feeding, lack of physical activities, exposure to radiations, and various forms of pollution, humans are prone to a deadly disease called cancer. Among different categories of cancer, breast cancer is affecting the women community very severely. Hence, the prediction of breast cancer in its earlier stage has become a crucial one in order to prevent from the loss of lives. At this juncture, various Machine Learning (ML) models can be trained for the diagnosis and prognosis of such disease. Nevertheless, owing to the presence of missing values and inefficient features in the dataset, the prediction accuracy of those models will get reduced. Hence, this paper strives to show that when ML classification algorithms are trained with datasets that are properly pre-processed with imputation of unknown values and selection of suitable features using feature selection techniques, the prediction accuracy can be improved.The experimental results on breast cancer datasets demonstrate the effectiveness by means of using KNN imputation method and feature engineering with feature selection using Recursive Feature Elimination(RFE) and Correlation based Feature Selection(CFS) method with 94.63% and 98.05% of accuracy. It could be concluded from the experimental results that the proposed approach can also be used for other medical diagnosis problems.