This dashboard monitors a model for traffic crashes in Chicago that predicts whether a crash involves an injury or not. Learn more about this model monitoring project here:
https://github.com/juliasilge/modelops-playground
The dashed lines correspond to the model metrics (accuracy and ROC AUC) measured on testing data when the model was last trained, 538 days ago.
This plot shows the new crashes used for monitoring since the last time this dashboard was generated.
This data exhibits class imbalance, which was addressed during modeling via downsampling.
This table shows the new crashes that our model did the worst job at classifying, in either direction.
Sort and filter to explore these misclassifications.
This dataset covers traffic crashes on city streets within Chicago city limits under the jurisdiction of the Chicago Police Department.
This map, like the previous panels, shows the new crashes used for monitoring since the last time this dashboard was generated.
Interact directly with the model for crashes via its visual documentation, and get curl examples.