Predicting hospital readmission of diabetics using deep forest

Pengwei Hu, Shaochun Li, Yu An Huang, Lun Hu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

16 Scopus citations

Abstract

Diabetes can cause a variety of complications, which also leads to a high rate of repeated admission of patients with diabetes, which greatly increases the pain and financial burden of patients. Higher readmission rates also reduce hospital evaluation and operational efficiency. Therefore, it is urgent to screen out high-risk readmission patients in advance and introduce adjuvant treatment to reduce the probability of readmission. In this study, we propose a deep learning model combining wavelet transform and deep forest to hospital readmission of the diabetic. The proposed model has been tested with real clinical records and compared with several prevalent approaches to patient prediction. The experimental results show that the feature representation transformed by wavelet transform may well represent the original features and the deep forest is able to outperform the state-of-the-art approaches to classify diabetics.

Original languageEnglish
Title of host publication2019 IEEE International Conference on Healthcare Informatics, ICHI 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781538691380
DOIs
StatePublished - Jun 2019
Externally publishedYes
Event7th IEEE International Conference on Healthcare Informatics, ICHI 2019 - Xi'an, China
Duration: 10 Jun 201913 Jun 2019

Publication series

Name2019 IEEE International Conference on Healthcare Informatics, ICHI 2019

Conference

Conference7th IEEE International Conference on Healthcare Informatics, ICHI 2019
Country/TerritoryChina
CityXi'an
Period10/06/1913/06/19

Keywords

  • Deep forest
  • Diabetes
  • Feature engineering
  • Hospital readmission

Fingerprint

Dive into the research topics of 'Predicting hospital readmission of diabetics using deep forest'. Together they form a unique fingerprint.

Cite this