Document Type
Student Coursework
Publication Date
2025
Abstract
Hospital readmissions, particularly among diabetic patients, place a significant burden on healthcare systems by increasing operational costs and straining limited resources. This project presents a scalable, cloud-based solution that leverages machine learning and big data analytics to predict 30-day hospital readmissions. Utilizing a ten-year dataset of over 100,000 patient records, we implemented a Random Forest classifier trained on clinical, demographic, and hospitalization data. The system architecture integrates Google Cloud Platform services—including BigQuery, Vertex AI, and Looker Studio—with a custom Python/Flask web application for real-time data input and inference. Data preprocessing and feature engineering were conducted in Vertex AI Workbench, enabling the transformation of raw medical records into model-ready formats. Our deployed model achieves high accuracy and supports prediction through a REST API endpoint, with interactive dashboards providing actionable insights to healthcare providers. The project demonstrates the potential of artificial intelligence to support proactive care management and reduce hospital readmissions, while laying the groundwork for automated retraining pipelines to accommodate evolving patient data.
Program or Discipline Name
Computer and Information Sciences
Recommended Citation
Ike, R., Kim, M., Chindarkar, B., & Cha, S. (2025). Readmission Prediction for Diabetic Patients: A Scalable Big Data Approach for Resource-Constrained Hospitals. Readmission Prediction for Diabetic Patients: A Scalable Big Data Approach for Resource-Constrained Hospitals, 1-4. Retrieved from https://digitalcommons.harrisburgu.edu/other-works/12
Publication Title
Readmission Prediction for Diabetic Patients: A Scalable Big Data Approach for Resource-Constrained Hospitals
Start Page No.
1
End Page No.
4