About
A combination of publicly available data sources with community, individual, social and opioid crisis related variables can be used to predict where treatment facilities may be needed in the future throughout the state of Indiana. To complete this task, we did the following:
- Collected and cleaned raw, public datasets.
- Created index variables.
- Developed and trained machine learning models by utilizing multinomial, spatial classification model with regularization to predict geolocation of high/medium/low opioid treatment facilities with aforementioned three latent variables as well as spatial variables (county area and size as well as proximity to existing treatment centers).
- Cross-validated models with held-out data.
Partners
Funding
- BioCrossroads
Publications and presentations