Predictive System of Traffic Congestion based on Machine Learning
Chaimaa Chaoura, Hajar Lazar, Zahi Jarir
Abstract
Growing traffic demand at junctions causes major traffic congestion, which can have a detrimental impact on the economy, health, and other aspects. This necessitates intelligent traffic management, which makes traffic prediction a vital task for intelligent transportation systems. We propose a method based on machine learning algorithms capable of predicting traffic flow at intersections, in which we split our dataset according to the four junctions, also we normalize and differentiate these data, then we implemented 11 algorithms based on Recurrent Neural Networks (GRU), Extra Trees Regressor (ET), Light Gradient Boosting Machine (LIGHTGBM), Random Forest Regressor (RF), Linear Regression (LR), Bayesian Ridge (BR), Gradient Boosting Regressor (GBR), K Neighbors Regressor (KNN), Decision Tree Regressor (DT), Huber Regressor (HUBER), Ridge Regression (RIDGE). Finally, we used two measures, RMSE and MAE, to evaluate these models in order to select the most efficient model for each part of the data set.