United States Department of Transportation
i
Collaborative Research Project to Coordinate the Data from the CRASH Predictive Analytics Program Between TDOT and TDOSHS
-
2021-09-01
[PDF-3.72 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report, December 2019-July 2021
-
Corporate Publisher:
-
Abstract:Emergency Response Management (ERM) necessitates the use of models capable of predicting the spatial-temporal likelihood of incident occurrence. These models are used for proactive stationing in order to reduce overall response time. Traditional methods simply aggregate past incidents over space and time; such approaches fail to make useful short-term predictions when the spatial region is large and focused on fine-grained spatial entities like interstate highway networks. This is partially due to the sparsity of incidents with respect to space and time. Furthermore, accidents are influenced by several risk factors that must be considered in predictive models. Collecting, cleaning, and managing multiple streams of data from various sources is challenging for large spatial areas. In this report, we highlight how this problem is being solved in collaboration with TDOT to improve ERM in TN. Working with TDOT, we have developed a novel pipeline to forecast road accidents on the interstate networks. Our pipeline, based on a combination of synthetic resampling, clustering, and data mining techniques, can efficiently forecast the spatial-temporal dynamics of accident occurrence, even under sparse conditions. Our pipeline uses data related to roadway geometry, weather, historical accidents, and traffic to aid accident forecasting. To understand how our forecasting model can affect allocation and dispatch, we improve and employ a classical resource allocation approach. Experimental results show that if our approach for forecasting road accident likelihood is employed in proactive ERM, it can reduce response times(up to19%for 20 availableHELPtrucksandupto8%onaverageformultipledifferentnumbersofavailableHELPtrucks)and the number of unattended incidents(up to75% and 50% for mean number and maximum number of unattended accidents during a 4-hour time window, respectively)in comparison to current approaches followed by first responders. The developed pipeline is efficacious, applicable in practice, and open source. Our feature analysis also showed combination of congestion and heavy rain can increase the rate of incidents by a factor of seven while visibility and wind speed do not playa key role in prediction of the likelihood of incidents. Finally, albeit significant improvement, we provide recommendations and techniques, which we aim to investigate and apply during the next phase of the project, to enhance the performance of the model even further. The pipeline is available onhttps://tn.statresp.ai/.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
