Urban Road Traffic Anomaly Detection Based On Isolation Forest

Urban road traffic anomaly detection is an important part of intelligent transportation system,detects abnormal roads with data from fixed detectors or travel floating car,provides advice for urban traffic department to carry management or planning.Orienting to the problem of detecting urban road anomaly traffic,this thesis suggests urban road traffic anomaly detection based on Isolation Forest.Details are summarized as following:(1)Simulation of traffic data and real traffic data processing.This paper not only uses the simulation software to collect the fixed detector data,but also the real traffic data processing,the Internet will be open to the electronic map data using PHP language to read to the MYSQL database,the extraction of Beijing’s four-ring urban traffic network.(2)Based on Isolation Forest detection of urban traffic with traffic anomaly.First improve the point-to-line map matching algorithm,the floating car data to match the electronic map,so that the floating car to match the right path to its driving.And then use the map data after the map to calculate the road traffic characteristics,according to the real traffic data to build Isolation Forest training data set,and finally use the trained Isolation Forest to detect traffic anomalies occurred in urban roads.While training Isolation Forest on the simulation data to detect anomalies in the road.(3)Improve the heat diffusion model to infer origin of anomaly traffic.When the road in the urban road network is abnormal,it will spread to the adjacent road,this phenomenon is similar to the heat diffusion phenomenon,so the heat diffusion model can be applied to the urban traffic anomaly source inference.This paper considers the similarity of traffic conditions among adjacent roads,rather than simply thinking that the anomalous transfer coefficients between all adjacent roads are the same.