Motivation
Traffic flow prediction is the key component of Intelligent Transport Systems (ITS)and can assist ITS to forecast and prevent traffic congestion, control and manage traffic efficiently, and plan the best traveling route. Data-driven traffic flow prediction has received extensive attention recently due to the availability of massive traffic data. Machine learning (ML) based approaches such as k-nearest neighbor (kNN) algorithm and Artificial Neural Network (ANN) schemes have been used. Compared with conventional ML methods, deep learning models have the advantages such as simplifying data preprocessing procedure and outperforming other ML methods in terms of accuracy. Therefore, deep learning schemes have received extensive attention recently in traffic flow prediction.
The objective is to study deep learning methods for traffic flow prediction and investigate the advantages/disadvantages of using deep learning methods, comparing with the traditional approaches.
Goal
To develop and validate deep learning methods for traffic flow prediction
What will you do?
- Good understanding of the traffic flow prediction problem
- Survey of existing prediction methods for traffic flow
- Reproduce results in the reference [1]; and compare with the results with previous machine learning methods
- Validate the developed algorithms and model in real dataset in Norway, if available
- Python or R programming language can be use
Supervisors
- Yan Zhang, IFI, University of Oslo, Norway
- Frank Eliassen, IFI, University of Oslo, Norway
Reference
[1] Traffic flow prediction with big data: a deep learning approach, IEEE Transaction on ITS, april 2015