Members of the Data Science Working Group at Code for San Francisco have been charged with answering the Community Technology Alliance’s prompt about homelessness programs.
What variables best predict whether an individual is categorized as ‘in permanent housing’ as an outcome, by population segment:
- Veterans
- Chronically Homeless
- Continuously Homeless
- Has Disabling Condition
- Domestic Violence Victim
- Male/Female
- Latino/Non-Latino
Data is in HMIS format, a data standard defined by the US Department of Housing and Urban Development
View the HMIS Data Science Study Presentation for a summary of our findings
- Load and clean the data
- Explore the data
- Feature engineering to prepare input variables
- Make outcome predictions with logistic regression model
Install Jupyter Notebook; this is most easily done by installing Anaconda: https://www.continuum.io/downloads
Install seaborn. To do this in a new conda environment:
conda create --name datasci seaborn
To deactivate/activate the environment:
source deactivate datasci
source activate datasci
- Fork this repository and clone it locally.
- Locate the dataset (pinned in #datasci-homeless on Slack).
- Run
jupyter notebook - Navigate to notebooks/load_data_example_v2.ipynb to start exploring the data.
Additional information on completed and open items can be found in the pinned documents in #datasci-homeless.