Short and long-term prediction of traffic flow using Machine-learning and Deep learning techniques..
Item Type:Conference Paper
Citation:Dhivyabharathi Bhaskaran, Bidisha Ghosh, Short and long-term prediction of traffic flow using Machine-learning and Deep learning techniques.., 14th International Conference on Applications of Statistics and Probability in Civil Engineering (ICASP14), Dublin, Ireland, 2023.
submission_399.pdf (PDF) 837.5Kb
Flow prediction is paramount for various intelligent transportation system applications in the traffic engineering domain. Prediction can be single-step ahead or multi-step ahead with each having its own set of advantages/limitations. However, predicting multiple steps ahead provides more insights about future traffic conditions/trends if the accuracy is not compromised. It is universally known, that featuring temporally farther inputs makes the model less robust, and hence, improving prediction accuracy is a challenge in long-term prediction. In this regard, the present study explored two different non-parametric techniques to perform multi-step ahead prediction of traffic flow using toll transaction data on Irelandﾒs prominent motorways. In this study, Support Vector Regression (SVR) and Long Short-Term Memory (LSTM) techniques are adopted to model the traffic flow time series and are implemented for long-term and short-term prediction. The results show that the performance of LSTM is found to be better than the SVM in the majority of the cases. The experiments on real traffic data show the advantages of deep learning models, demonstrating the potential and promising capability of the proposed framework for multi-step traffic flow prediction.
Other Titles:14th International Conference on Applications of Statistics and Probability in Civil Engineering(ICASP14)
Type of material:Conference Paper
Series/Report no:14th International Conference on Applications of Statistics and Probability in Civil Engineering(ICASP14)
Availability:Full text available