Deep multilayer and nonlinear Kernelized Lasso feature learning for healthcare in big data environment
- PDF / 1,338,180 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 80 Downloads / 171 Views
ORIGINAL RESEARCH
Deep multilayer and nonlinear Kernelized Lasso feature learning for healthcare in big data environment S. Prakash1 · K. Sangeetha2 Received: 17 April 2020 / Accepted: 10 July 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract In this modern era, healthcare industry is being metamorphosed by the progress in machine learning (ML). By utilizing vast big data, ML is now being pre-owned in healthcare to bestow comparatively better patient care and has emerged in enhanced business consequences. In this paper an effective processing framework called deep multilayer and non-linear Kernelized Lasso feature learning (DM-NKLFL) is introduced to powerfully cope with the data explosion in image processing field. Our work dedicates to provide a general framework for both simple linear and complex non-linear relationships. This in turn helps to handle the increase in image scale without affecting the performance. The proposed DM-NKLFL method includes two parts, i.e., stepwise regression nonlinear Kernelized Lasso (SR-NKL) feature selection and deep multilayer pattern learning (DMPL). Specifically, SR-NKL is aimed at processing non-linear features to minimize time and complexity involved during feature selection whereas the DMPL is proposed to deeply learn data driven features to determine the underlying patterns. The DM-NKLFL method over the traditional state-of-the-art methods are validated both in time efficiency and quality of results using the big biological data. Keywords Machine learning · Deep multilayer · Non-linear · Kernel · Lasso feature learning
1 Introduction Health Informatics field is on the upsurge, entering a new era where technology has starting to handle Big Data, nurturing unlimited possibilities for information growth. Both data mining and Big Data analytics are assisting to capitalizing the goals of diagnosing, detection and providing remedies to patients in need of healthcare. In the recent years, computed tomography (CT) is extensively utilized to help disease diagnosis. However, it is a time consuming process for assisting early lung cancer diagnosis, from a large number of pulmonary CT images. Several machine learning methods homogenized with image processing procedures were exploited with objective * S. Prakash [email protected] K. Sangeetha [email protected] 1
CSE Department, Sri Shakthi Institute of Engineering and Technology, Coimbatore, India
CSE Department, SNS College of Technology, Coimbatore, India
2
of automating the selection and validation of the vast image data. In Ding et al. (2017), deep learning based image data selection was utilized where machine learning based techniques consisting of support vector machine (SVM) and deep learning were integrated for automating the process of data selection. Besides, an experiment based approach was also designed with the purpose of optimizing the features to enhance the rate of accuracy of machine learning based classification. For measuring the machine learning algorithms, both cr
Data Loading...
