Curvature-based feature selection with application in classifying electronic health records

Disruptive technologies provides unparalleled opportunities to contribute the identifications of many aspects in pervasive healthcare, from adoption Internet Things through Machine Learning (ML) techniques. As a powerful tool, ML has been widely applied patient-centric healthcare solutions. To further improve quality patient care, Electronic Health Records (EHRs) are commonly adopted facilities for analysis. It is crucial task apply AI and analyse those EHRs prediction diagnostics due their highly unstructured, unbalanced, incomplete, high-dimensional nature. Dimensionality reduction common data preprocessing technique cope with EHR data, which aims reduce number features representation while improving performance subsequent analysis, e.g. classification. In this work, an efficient filter-based feature selection method, namely Curvature-based Feature Selection (CFS), presented. The proposed CFS concept Menger Curvature rank weights all given set. evaluated four well-known sets, including Cervical Cancer Risk Factors (CCRFDS), Breast Coimbra (BCCDS), Tissue (BTDS), Diabetic Retinopathy Debrecen (DRDDS). experimental results show that achieved state-of-the-art on above sets against conventional PCA other most recent approaches. source code approach publicly available at https://github.com/zhemingzuo/CFS.