Forecasting Demand for Urgent Care Services
Jul. 2023 – Feb. 2024
Machine learning based forecasting for healthcare demand prediction, collaborated with BorealisAI.
Overview
This team project was conducted from July 2023 to February 2024 in collaboration with Borealis AI and the University of British Columbia. The goal was to forecast urgent care and emergency department demand more accurately, with the broader aim of supporting healthcare resource allocation and staffing decisions.
The project explored whether machine learning methods could improve short term forecasting performance compared with conventional statistical baselines.
Project Scope
The work began with the Hospital Triage Dataset from Kaggle and extended it by integrating external contextual variables, including local weather conditions and holiday information. These additional features were incorporated to better capture demand patterns that may not be reflected in the original hospital data alone.
A range of forecasting methods were applied and evaluated, including neural networks and time series models. Model performance was then compared against baseline statistical approaches.
Contributions
-
Applied multiple machine learning and time series forecasting models to predict local emergency department demand.
-
Built an enriched prediction pipeline by combining hospital triage data with climate and holiday features.
-
Evaluated forecasting accuracy against baseline methods and identified settings in which machine learning models provided stronger short term predictions.
Outcome
The project achieved noticeable improvements in short term forecasting accuracy over baseline statistical models. The results highlighted the value of data driven forecasting for improving operational planning in healthcare systems, especially for staffing and service capacity management.
Repository
The implementation and related project materials are available here: