Real-Time Identification of Risk-Prone Traffic Patterns Taking into Account Weather Conditions
This paper presents the concept of a methodology for identifying in real time risky motorway traffic conditions. This research effort is to automate the extraction of traffic patterns from large data archives using advanced computing techniques. These patterns are impacted by interferences such as crashes, work zones and weather conditions. The traffic data in an archive can be analyzed as a time series signal, consisting of an underlying pattern and noise. Such a perspective allows the application of advanced computing techniques like clustering, fuzzy logic and other classification algorithms to separate the background traffic pattern from the noise. Self-Organizing Maps (SOM) has emerged as a promising tool in the recent past for clustering field. In this paper, PCA transform with SOM tool are used to extract underlying traffic patterns and identification of weather-sensitive and risk-prone traffic patterns. Each group can be considered as a traffic regime containing similar traffic situations. The applicability of the methodology in term of active traffic management is also discussed.