Crashmap360: a data-driven infrastructure recommendation system for Washington, D.C.

Traffic safety remains a persistent challenge in Washington, D.C., despite initiatives such as Vision Zero and The Lab @ D.C. Rather than functioning as a general crash-prediction model, CrashMap360 is a decision-support framework that identifies candidate locations for infrastructure review by integrating crash severity, infrastructure proximity, traffic exposure, and spatial clustering. The framework was applied to three crash categories: nighttime crashes, speeding-related crashes, and cyclist-involved crashes. Within each category, crashes were filtered by type, distances to the nearest relevant safety infrastructure were calculated, and crashes beyond predefined infrastructure-distance thresholds were clustered and prioritized using crash frequency, severity-based metrics, and exposure-normalized risk. The robustness of the framework was evaluated through comparisons with capped-severity rankings, crash-frequency rankings, exposure-normalized rankings, DBSCAN clustering, the District Department of Transportation’s High Injury Network, and threshold sensitivity analyses. The results identified candidate locations for engineering review across all three crash categories and showed that hotspot prioritization depends on the ranking methodology, with exposure normalization and severity weighting producing category-specific changes in prioritization while threshold perturbations had relatively little effect on the ranking of locations. These findings suggest that CrashMap360 provides a robust and flexible framework for systematically prioritizing candidate locations for engineering review while complementing existing transportation safety analyses.