Abstract: Due to the characteristics of high driving speed and difficulties of controlling traffic flow, highway traffic accidents are often more serious than that of urban roads. Hence, it is a crucial issue in the traffic engineering field to identify the highway black spots and analyze their spatio-temporal association patterns. Traditional statistical methods may be confronted with the problem of modifiable areal unit problem(MAUP), which means the accuracy of identifying accident black spots directly depends on the basic unit size. Traffic accident is a representative spatio-temporal event, which often contains particular spatio-temporal patterns. The paper proposes a novel method to identify the highway black spots based on spatio-temporal density-based spatial clustering of applications with noise (DBSCAN) clustering. The proposed method considers the time periodicity and accident severity into calculating the spatio-temporal neighboring indicators and then finds various traffic accident black spots of multi-spatio-temporal scales based on the density-connecting rules of DBSCAN. The experimental results of 2012-2016 highway crash datasets in Hunan province illustrate the proposed method can adaptively identify the traffic black spots of different length, efficiently overcome MAUP problem, and meanwhile mine the spatio-temporal aggregative characteristics of traffic accidents to provide auxiliary decision supports for making suitable emergency plans for traffic safety.

Key words: traffic accidents analysis, black spots road, density clustering, MAUP, spatio-temporal characteristics