Multi-View Actionable Patterns for Managing Traffic Bottleneck

Publisher:
AAAI Press
Publication Type:
Conference Proceeding
Citation:
Artificial Intelligence for Transportation: Advice, Interactivity and Actor Modeling: Papers from the 2015 AAAI Workshop, 2015, pp. 64 - 70
Issue Date:
2015
Full metadata record
Files in This Item:
Filename Description Size
art.pdfPublished version523.37 kB
Adobe PDF
Discovering congestion patterns from table-formed traf- fic reports is critical for traffic bottleneck analysis. However, patterns mined by existing algorithms often do not satisfy user requirements and are not actionable for traffic management. Traffic officers may not pursue the most frequent patterns but expect mining outcomes showing the dependence between congestion and various kinds of road properties for traffic planning. Such multi-view analysis requires to integrate user preferences of data attributes into pattern mining process. To tackle this problem, we propose a multi-view attributes reduction model for discovering the patterns of user interests, in which user views are interpreted as preferred attributes and formulated by attribute orders. Based on the pattern discovery model, a workflow is built for traf- fic bottleneck analysis, which consists of data preprocessing, preference representation and congestion pattern mining. Our approach is validated on the reports of road conditions from Shanghai, which shows that the resultant multi-view findings are effective for analyzing congestion causes and traffic management.
Please use this identifier to cite or link to this item: