Development of Big Data Classifier for Biomedicine Early Diagnosis: An Experimental Approach Using Machine Learning Methods

Abstract

In the fast-phase world, data availability is abundant due to a rapid adaptation increase of big data technologies. Large amounts of data have been generated and collected at an unprecedented speed and scale, introducing a revolution in medical research practices for biomedicine informatics. Thus, there is an immense demand for statistically rigorous approaches, especially in the medical diagnosis discipline. Therefore, this study utilized the Bayesian Belief Network (BBN) for feature selection, which identifies relevant features from a larger set of attributes and employs a stratification for the Stochastic Gradient Descent (SGD) classifier in the classifying of breast cancer on the publicly available machine learning repository at the University of California, Irvine (UCI) such, breast cancer Wisconsin and Coimbra breast cancer datasets. The experimental approach of using BBN as feature selection achieved 0.95% coincidence. Thus, a stratified Stochastic Gradient Descent (SGD) was employed to build a classification model to validate the coincidence. Our proposed modeling classifier approach reached novelty 98%, which improved by 7% compared to the previous works. Furthermore, this study presents a web-based application, a prototype type, to employ the proposed classifier model for breast cancer diagnosis. This study expects to provide a source of confidence and satisfaction for medical physicians to use decision-support tools.