Predicting Ambulance Call, Incident and Response data by hour or by day over a 6 week to 1 year forecast horizon
This project involved building a forecasting model for the South West Ambulance Services Trust.
What was created was “a bespoke forecasting model, capable of predicting Call, Incident and Response data by hour for a period of up to 6 weeks or daily over 365 days. The Forecasting & Capacity Planning Team at SWASFT then used Python and SQL computer modelling software to convert the forecast models into useable rota patterns. The project has resulted in a fully automated, accurate and safe demand-led resourcing model.”
The work used an ensemble of different time series forecasts with a proven ability to match real-world patterns.
https://arc-swp.nihr.ac.uk/news/data-science-ambulance-demand/
The code is available here: https://github.com/TomMonks/swast-benchmarking