Accident Risk Prediction based on Heterogeneous Sparse Data: New Dataset and Insights

Published in The 27th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL 2019). (Chicago, IL), 2019


Sobhan Moosavi, Mohammad Hossein Samavatian, Srinivasan Parthasarathy, Radu Teodorescu, Rajiv Ramnath

Summary and Download

This paper is a new solution for real-time traffic accident prediction based on heterogeneous data such as traffic, weather, and points-of-interest. It also provides a new traffic accident dataset which is called the US Accidents. The outline of our proposed Deep Accident Prediction (DAP) model is presented below:

Download paper here

Access code here

Download slides presented at SIGSPATIAL here