Modeling and forecasting vehicle flows at a roundabout using wavelets and neural networks

Moussa Koné; Vincent Monsan; Sylvestre Placide Ekra

doi:10.56947/amcs.v32.711

Authors

Moussa Koné UFR Mathematiques et Informatique, Universite Felix Houphouet Boigny, Cote d Ivoire
Vincent Monsan UFR Mathématiques et Informatique, Université Félix Houphouët Boigny, Côte d’Ivoire
Sylvestre Placide Ekra Unit´e de Recherche et d’Expertise Num´erique, Universit´e Virtuelle de Cˆote d’Ivoire, Cˆote d’Ivoire

DOI:

https://doi.org/10.56947/amcs.v32.711

Keywords:

Traffic forecasting, Finite mixture model, Wavelet thresholding, LSTM neural networks, Traffic density estimation

Abstract

We propose a real-time traffic forecasting method for urban intersections, using a finite mixture model and wavelet-based density estimation. Our approach models traffic flow from four directions using high-frequency data. The weights of the finite mixture model components are predicted by an LSTM neural network. Results demonstrate good predictive accuracy, with low MAE and stable KL divergence.