*Thanks to Terry Zdan for the share...*

*Data Mining to Improve Planning for Pedestrian and Bicyclist Safety* https://www.vtti.vt.edu/utc/safe-d/wp-content/uploads/2019/12/01-003_Final-R...

*Abstract* Between 2009 and 2016, the number of pedestrian and bicyclist fatalities saw a marked trend upward. Taken together, the overall percentage of pedestrian and bicycle crashes now accounts for 18% of total roadway fatalities, up from 13% only a decade ago. Technological advancements in transportation have created unique opportunities to explore and investigate new sources of data for the purpose of improving safety planning. This study investigated data from multiple sources, including automated pedestrian and bicycle counters, video cameras, crash databases, and GPS/mobile applications, to inform bicycle and pedestrian safety improvements. Data mining techniques, a new sampling strategy, and automated video processing methods were adopted to demonstrate a holistic approach that can be applied to identify facilities with highest need of improvement. To estimate pedestrian and bicyclist counts at intersections, exposure models were developed incorporating explanatory variables from a broad spectrum of data sources. Intersection-related crashes and estimated exposure were used to quantify risk, enabling identification of high-risk signalized intersections for walking and bicycling. The modeling framework and data sources used in this study will be beneficial in conducting future analyses for other facility types, such as roadway segments, and also at more aggregate levels, such as traffic analysis zones.

*Conclusions and Recommendations* While statistics show an increasing trend in using eco-friendly modes of travel, such as walking and bicycling, historical crash data shows a growing trend in road crash victims involving pedestrians and bicyclists. Utilizing multiple data sources, such as automated pedestrian and bicycle counters and video cameras, this study estimated pedestrian and bicyclist exposure and identified signalized intersections with the highest risk for walking and bicycling within the city of San Diego, California.

A sampling strategy was used to identify a representative sample of intersections to collect short-term video data by applying cluster analysis and stratified sampling. A vision-based monitoring system was developed to automatically detect, track, and count pedestrians and bicyclists at the selected intersections. Situations where high-quality videos were used, a sufficient number of pedestrians and bicyclists were annotated, pedestrians and bicyclists were not too far from the camera, did not cross the intersection in groups, and good lighting was present, led to a high counting accuracy of 95%. Utilizing permanent counters, an extrapolation and a novel matching method were employed to estimate yearly counts that were used for estimating exposure by direct demand models. Exposure analysis identified transportation network, population, traffic generator, and land use variables as statistically significant in estimating pedestrian and bicyclist volume. Accounting for exposure as a normalization factor and considering other factors, such as frequency of victims and crash severity, in quantifying risk had a significant impact on the selection of high-risk intersections; not all intersections with the highest number of pedestrian and bicyclist victims were identified as high-risk. In addition, the variables were found to be influential at multiple buffer areas and showed differences across pedestrian and bicycle activity. The results underscored the importance of location and community in characterizing non-motorized demand, and targeted improvements to encourage non-motorized activities.

The modeling framework and data sources used in this study will be beneficial in conducting future analyses for other facility types, such as roadway segments, and also at more aggregate levels, such as traffic analysis zones. The approach is also beneficial to public agencies, as it can help quantify the risk of walking and bicycling at intersections, which in turn can aid in the development of procedures to identify high-risk facilities and prioritize them for countermeasure implementation. It should be pointed out that safety performance functions were not considered in this study, and thus potential future work might focus on how a combination of risk quantification and performance functions can assess safety. Since crashes are rare events, the identification of high-risk facilities will be lengthy, and a potential future direction is to proactively assess safety by discovering near-crash situations in video analysis. This would enable researchers and practitioners to quantify risk and evaluate safety in a much shorter period of time.