Exploring Crowdsourced Big Data to Estimate Border Crossing Times

This project provides an in-depth assessment of market-available connected vehicle (CV) data--Wejo CV data in border crossing time estimation. This project evaluates both historical and streaming Wejo CV data collected at the Paso del Norte (PDN) port of entry (POE) in El Paso. The research team first developed a set of big data analytic tools to process Wejo datasets and generated CV-based border crossing times (CV-Time). Then, the research team evaluated the temporal coverage of CV-Time at the PDN POE and its correlation with the existing Bluetooth-generated border crossing times (Bluetooth-Time). The results demonstrate that the CV-Time is strongly correlated with the Bluetooth-Time with a correlation rate of approximately 0.8. The best fitted ordinary least squares model based on the combination of CV-Time and additional CV-generated variables can produce a temporally transferable model to estimate Bluetooth-Time with an R2 of 0.88, root mean square error (RMSE) of about 15 min and mean absolute percentage error (MAPE) of 52%. Random forest and gradient boosting models were also investigated to decrease the MAPE. The research findings show that the gradient boosting has the best performance among the introduced models with an RMSE of 15.5 min and MAPE of 26%. However, this evaluation was conducted based on the 2022 Wejo data with a relatively low penetration rate, resulting in around 30 percent of the testing hours lacking Wejo samples. Given these findings, the Wejo CV data can be a potential and promising data source for monitoring border crossing times, especially as sample penetration rates improve. The Wejo CV data can be used to supplement the existing Border Crossing Information System (BCIS) in El Paso and could be treated as an alternative solution when the BCIS is down for maintenance.