Time Series Prediction Using Deep Learning Methods in Healthcare

Traditional machine learning methods face unique challenges when applied to healthcare predictive analytics. The high-dimensional nature of data necessitates labor-intensive and time-consuming processes selecting an appropriate set features for each new task. Furthermore, depend heavily on feature engineering capture the sequential patient data, oftentimes failing adequately leverage temporal patterns medical events their dependencies. In contrast, recent deep (DL) have shown promising performance various prediction tasks by specifically addressing data. DL techniques excel at useful representations concepts clinical as well nonlinear interactions from raw or minimally processed this article, we systematically reviewed research works that focused advancing neural networks structured time series tasks. To identify relevant studies, searched MEDLINE, IEEE, Scopus, ACM Digital Library publications through November 4, 2021. Overall, found researchers contributed literature in 10 identifiable streams: models, missing value handling, irregularity, representation, static inclusion, attention mechanisms, interpretation, incorporation ontologies, strategies, scalability. This study summarizes insights these streams, identifies several critical gaps, suggests future opportunities applications using