Skip to content

Latest commit

 

History

History
34 lines (24 loc) · 4.01 KB

README.md

File metadata and controls

34 lines (24 loc) · 4.01 KB

Dallas Data Challenge

This repository contains two Machine Learning models that forecast the daily # of available and occupied hospital beds in the Dallas-Fort Worth region based on previous data.

That is, when given future dates as input, the model will forecast the # of available hospital beds on each of those future dates and in addition, provide a range of possible values.

About the Code

The code is easily modifiable and very detailed so that predictions can be made for different cities/regions given their data.

Features of the Notebooks

  • Evaluating Model Performance on Known Data:
    • In each Jupyter Notebook I train the model on data only before September - (data is available for September dates).
    • After the model is trained, I make it "forecast" the # of available/occupied hospital beds in the first 3 weeks of September.
    • I compare these forecasts with the actual # of hospital beds and calculate the percent error, mean absolute error, and # of forecast intervals that captured the correct # of beds)
  • Forecasting for Future Dates:
    • In this part of the notebook, I train the model on all past hospital bed data (including September), then generate a list of 31 days from September 23rd to October 23rd and feed it into the model.
    • The model forecasts the # of available/occupied hospital beds in the DFW region for future dates (late September - October).

Features of the Model

  • Providing a forecast of the # of hospital beds available/occupied for a specified date in the future.
  • Providing an interval of possible # of hospital beds for a specified date in the future.

About the Data

The data comes from the Texas DSHS website: https://dshs.texas.gov/coronavirus/additionaldata.aspx

The CSVs in this repository are: