Traffic Modelling Using Convolutional Long-Short Term Memory Neural Network

Nicolae FLORIAN, Dumitru POPESCU, Severus-Constantin OLTEANU

Abstract

This paper examines hybrid neural networks for traffic management and prediction. These networks combine recurrent and convolutional neural network architectures. The current studies on FNN, RNN, and CNN architectures in this field are reviewed and the conventional traffic modelling techniques and their limitations are also evaluated. The proposed hybrid model integrates the strengths of two architectures: it uses RNNs for identifying the temporal dependencies in traffic data and it employs CNNs for recognizing spatial patterns. The paper details the employed methodology, which includes data collection, network architecture design, training, hyperparameter tuning, and model performance evaluation. The implications of the findings of this paper for urban planning and intelligent transportation systems (ITS) are discussed and the obtained results are compared with those achieved by the conventional approaches.

Keywords

Traffic, Modelling, Neural network, Convolution, Prediction, Intelligent transportation system.

View full article