Road traffic prediction using context-aware random forest based on volatility nature of traffic flows

Authors:
Narjes Zarei;Mohammad Ali Ghayour;Sattar Hashemi
Affiliations:
Department of Computer Science and Engineering, Shiraz University, Shiraz, Iran;Department of Computer Science and Engineering, Shiraz University, Shiraz, Iran;Department of Computer Science and Engineering, Shiraz University, Shiraz, Iran
Venue:
ACIIDS'13 Proceedings of the 5th Asian conference on Intelligent Information and Database Systems - Volume Part I
Year:
2013

Citing 6
Cited 0

Random Forests

Machine Learning
Multivariate short-term traffic flow forecasting using time-series analysis

IEEE Transactions on Intelligent Transportation Systems
A Convex Combination of Models for Predicting Road Traffic

ICDMW '10 Proceedings of the 2010 IEEE International Conference on Data Mining Workshops
Predicting Future Traffic Congestion from Automated Traffic Recorder Readings with an Ensemble of Random Forests

ICDMW '10 Proceedings of the 2010 IEEE International Conference on Data Mining Workshops
IEEE ICDM 2010 Contest: TomTom Traffic Prediction for Intelligent GPS Navigation

ICDMW '10 Proceedings of the 2010 IEEE International Conference on Data Mining Workshops
Traffic Simulation Framework

UKSIM '12 Proceedings of the 2012 UKSim 14th International Conference on Modelling and Simulation

Quantified Score

Hi-index	0.00

Visualization

Abstract

Nowadays short-term traffic prediction is of great interest in Intelligent Transportation Systems (ITS). To come up with an effective prediction model, it is essential to consider the time-dependent volatility nature of traffic data. Inspired by this understanding, this paper explores the underlying trend of traffic flow to differentiate between peak and non-peak traffic periods, and finally makes use of this notion to train separate prediction model for each period effectively. It is worth mentioning that even if time associated with the traffic data is not given explicitly, the proposed approach will strive to identify different trends by exploring distribution of data. Once the data corresponding trends are determined, Random Forest as prediction model is well aware of data context, and hence, it has less chance of getting stuck in local optima. To show the effectiveness of our approach, several experiments are conducted on the data provided in the first task of 2010 IEEE International Competition on Data Mining (ICDM). Experimental results are promising due to the scalability of the proposed method compared to the results given by the top teams of the competition.