chendaniely / red-cross-2 Goto Github PK

Where should the American Red Cross go to install smoke alarms?

In 2015, DataKind DC, the American Red Cross, and Enigma worked together to create a Home-Fire Risk Score at the U.S. Census Tract Level. The team created models to predict the highest impact areas to go knock on doors to install smoke alarms. These models were developed using proprietary data from the American Red Cross, the American Community Survey, the American Housing Survey, and NFIRS. Since the Red Cross doesn't visit all areas of the USA, risk scores were imputed to non-surveyed areas. Results were displayed on the smoke signals map. Some work was also done to understand historical home fires, where did it happen, how deadly were they?

The final product was a Home Fire Risk Map.

Google Drive link:

https://drive.google.com/drive/folders/1jq6iQiYgzQZM_vS_k2oiDlvyB5u-ywwW?usp=sharing

1. Replicate Phase 1 Models
2. Cross Validation of Phase 1 Predictions with New data
3. Improve/Update Index and models. We have new NFIRS, Red Cross, and ACS data that we would like to incorporate. We would like to consider adding new types of data as well such as climate data.
4. Update the Home Fire Risk Map!

The first step of the Red Cross Phase 2 is to replicate and document Phase 1 results. Phase 1's original GitHub repo is here. There are six models that the team is working to replicate, if you are interested in helping out on one, the Data Corps contact person is below.

DataKind DataCorps brings together teams of pro bono data scientists with social change organizations on long-term projects that use data science to transform their work and their sector. We help organizations define their needs and discover what’s possible, then match them with a team that can translate those needs into data science problems and solve them with advanced analytics.

We are very proud to re-partner with the American Red Cross!