Big data analytics framework for disease classification using machine learning techniques

Abstract

Most people in the world brazen out with lung disease. The controversial nature of lung disease relies on diagnosis and consequences. With the expanding utilization of earlier diagnosis, lung disease treatment would turn out to be more imperative in this world. Medical data, on the other hand, is structured, semi-structured, and unstructured. Combining with the healthcare system, the medical data with a large amount of higher-dimensional data includes EHR (Electronic Health Records), in-patient and out-patients data, patient s history and clinical images. Datasets associated with lung disease accrued and pre-processed. Sputum images deploys map-reduce framework to sort similar nature of images such as eosinophilia, squamous carcinoma, bronchial mucus which varies in staining. To extract nucleus and cytoplasm with its optimal features, sputum images retrieved from map-reduce endures feature extraction, that are most massive a part of figuring out the lung cancer. In the next step of pre-processing, optimized feature selection strategies inclusive of chi-square test, MDL, Relief and symmetric uncertainty are deployed and compared in all features of all datasets which includes sputum images, lung cancers, and thoracic surgical treatment. Sooner or later, proposed model promoted symmetric uncertainty with Minimum Redundancy Maximum Relevance to retrieve correlated features and most relevant features among large units of functions and datasets. newline