Replica customers can now purchase access to new traffic datasets for specific geographies. The initial set of available traffic datasets available are: Annual Average Daily Traffic (AADT) Data and Hourly Auto Volume Profiles, Annual Speeds Data, Turning Movement Counts (TMCs)
HPMS Validation
When producing AADT, we exclude a random 10% of HPMS ground truth data as a "test set" This is data the model has not "seen before" and allows us to quantify model performance. Below are tables and charts summarizing the model’s performance on the 2021 nationwide “test set”.
The model performs strongly. Across roads of FRC 1-4, expected error for total AADT in 2021 is 12.2%. Expected error for single-unit and combination truck volumes are 19.9%
We also report the signed median percent error: this statistic tells us if the model is consistently over- or under-predicting. For the total counts, this number is close to 0 for all roads, indicating that the model outputs an unbiased estimate.
New York City Benchmark
Replica obtained 2021 traffic camera speed data for a subset of links in NYC from the City’s open data portal. This data spanned approximately 30 locations in the NYC metro area. Traffic camera speed data is analogous to our method, as camera record the average speed between two points on a roadway via the elapsed time between them.
Trans-Manhattan Expressway (I-95) Comparison
This image is a comparison between between NYC’s camera data (“SPEED”) and Replica’s quarter-hourly speed data (“replica_speed”) on the Trans-Manhattan Expressway (I-95) for a typical Tuesday.
The blue line is NYC’s camera data and the red line is Replica’s speeds data.
Roseville, California — Fall 2022
We compare ground truth data for Roseville, CA for Fall 2022 to Replica outputs at 60 intersections. The ground truth data was provided to Replica by the City of Roseville, and is publicly available.
Roseville Ground Truth vs. Replica, average midweek daily counts per intersection for Fall 2022
The R-squared value is 0.98.