Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/68540
Title: Improving traffic prediction based on weather information
Authors: Xiao, Li
Keywords: DRNTU::Engineering::Electrical and electronic engineering
Issue Date: 2016
Abstract: With the high-paced development in every field of urban civilization, transportation becomes a key industry which is directly linked with the healthy and normal operation of other industries. As one of the most thriving international metropolis, Singapore has unfolded a long-term plan called Smart Mobility 2030 in order to build a fully- developed Intelligent Transportation System (ITS) by 2030, which can enhance commuters’ travelling experience through real-time transportation information, dynamic route guiding, traffic planning and autonomous accident detection and prevention, bi the field of traffic information, traffic prediction plays a crucial role due to its significance to traffic planning. However, the traffic is a chaos system so that the prediction accuracy can be impacted by many factors like weather. This study firstly analyzed on the predictability of traffic in Singapore according to the characteristics of traffic and then analyzed in the impact of different rainfall intensity on Singapore traffic based on the large-scale real data collected from local governmental organizations. We research on different fundamental statistical prediction theories and finally selected Support Vector Machine (SVM) and build a predicting model with RBF kemel function incorporating both traffic and rainfall data. By comparison, SVM prediction without weather information is also carried out. Then a three-fold cross validation is implemented to avoid over-fitting. The final results show that only for some particular links, the weather information can bring about increase in prediction accuracy. Some reasoning analysis shows that this may due to diverse factors like the speed changing frequency, condition of collected data, and the low resolution of speed data.
URI: http://hdl.handle.net/10356/68540
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
Xiao_Li_2015.pdf
  Restricted Access
Main article9 MBAdobe PDFView/Open

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.