Full Citation
Title: Generation of synthetic datasets using weighted bayesian association rules in clinical world
Citation Type: Journal Article
Publication Year: 2022
ISBN:
ISSN: 25112112
DOI: 10.1007/S41870-022-01081-X/FIGURES/3
NSFID:
PMCID:
PMID:
Abstract: Machine learning applications in the clinical domain are speedily changing the health care sector with refinement in cost and quality of service. Still, clinical data privacy concerns a lot, limiting the revolution in the clinical care industry. Patient’s clinical information is exceptionally subtle and personally identifiable, like medical history, ongoing conditions, payment and credit card information; due to these private data, regulations like Health Insurance Portability and Accountability heavily protect patient’s medical data. Here synthetic patient data artificially generated can be the best solution to tackle the challenges of health care transformation. Researchers, medical institutes, and companies building artificial intelligence solutions always need clinical data to work upon so synthetic data can help them eliminate the above challenges. In this proposal, weighted bayesian association rules concepts are applied to generate realistic synthetic data while preserving the relationships among the data and validating the synthetic information using the multivariate relationships method and predictive techniques. University of California Irvine datasets of heart disease, breast cancer, and diabetes are used to generate synthetic datasets to build weighted Bayesian belief networks with promising results. The proposed idea will allow healthcare organizations to efficiently distribute synthetic clinical data to researchers, develop inferences, and create analytical tools without compromising data privacy.
Url: https://link.springer.com/article/10.1007/s41870-022-01081-x
User Submitted?: No
Authors: Kharya, Shweta; Soni, Sunita; Swarnkar, Tripti
Periodical (Full): International Journal of Information Technology
Issue: 6
Volume: 14
Pages: 3245-3251
Data Collections: IPUMS USA
Topics: Health, Methodology and Data Collection, Population Data Science
Countries: