Network-level accident-mapping: Distance based pattern matching using artificial neural network.

摘要: The objective of an accident-mapping algorithm is to snap traffic accidents onto the correct road segments. Assigning segments facilitate robustly carry out some key analyses in accident research including identification hot-spots, network-level risk mapping and segment-level modelling. Existing algorithms have severe limitations: (i) they are not easily ‘transferable’ as specific given datasets; (ii) do perform well all road-network environments such areas dense network; (iii) methods used addressing inaccuracies inherent type environment. purpose this paper develop a new based on common variables observed most databases (e.g. name type, direction vehicle movement before recorded location). challenges here to: method that takes into account uncertainties data underlying digital network data, accurately determine proportion inaccuracies, robust can be adapted for any set varying complexity. In order overcome these challenges, distance pattern-matching approach identify segment. This vectors containing feature values data. Since each does contribute equally towards segments, ANN using single-layer perceptron assist “learning” relative importance calculation hence link identification. performance developed was evaluated reference dataset from UK confirming accuracy much better than other methods.