Abstract

International audience; Good, efficient and reliable public transportation systems are of crucial importance for all major cities today. In this paper, we propose a concrete solution to a particular problem: improve the prediction of the bus arrival time at each bus stop station on a given itinerary, by taking to account global and local traffic contexts. The main principle consists of modeling the traffic data as an image structure, adapted for applying CNN deep neural networks. The results obtained shows that the proposed approach outperforms traditional machine learning techniques, such as OLS (Ordinary Least Squares) or SVR (Support Vector Regression) with different kernels (RBF or Polynomial), with more than 18% better accuracy prediction, while being computationally faster.

Abstract

International audience; In today's modern cities, mobility is of crucial importance, and public transportation is particularly concerned. The main objective is to propose solutions to a given, practical problem, which specifically concerns the bus arrival time at various bus stop stations, by taking to account local traffic conditions. We show that a global prediction approach, under some global macro-parameters (e.g., total number of vehicles or pedestrians) is not feasible. This observation leads us to the introduction of a finer granularity approach, where the traffic conditions are represented in terms of a traffic density matrix. Under this new paradigm, the experimental results obtained with both linear and neural networks (NN) approaches show promising prediction performances. Thus, the NN approach yields 24% more accurate prediction performances than a basic, linear regression

Abstract

Predicting traffic conditions from online route queries is a challenging task as there are many complicated interactions over the roads and crowds involved. In this paper, we intend to improve traffic prediction by appropriate integration of three kinds of implicit but essential factors encoded in auxiliary information. We do this within an encoder-decoder sequence learning framework that integrates the following data: 1) offline geographical and social attributes. For example, the geographical structure of roads or public social events such as national celebrations; 2) road intersection information. In general, traffic congestion occurs at major junctions; 3) online crowd queries. For example, when many online queries issued for the same destination due to a public performance, the traffic around the destination will potentially become heavier at this location after a while. Qualitative and quantitative experiments on a real-world dataset from Baidu have demonstrated the effectiveness of our framework. Comment: KDD 2018. The first two authors share equal contribution

Abstract

Short-term prediction of traffic flow is one of the most essential elements of all proactive traffic control systems. The aim of this paper is to provide a model based on neural networks (NNs) for multi-step-ahead traffic prediction. NNs’ dependency on parameter setting is the major challenge in using them as a predictor. Given the fact that the best combination of NN parameters results in the minimum error of predicted output, the main problem is NN optimization. So, it is viable to set the best combination of the parameters according to a specific traffic behavior. On the other hand, an automatic method—which is applicable in general cases—is strongly desired to set appropriate parameters for neural networks. This paper defines a self-adjusted NN using the non-dominated sorting genetic algorithm II (NSGA-II) as a multi-objective optimizer for short-term prediction. NSGA-II is used to optimize the number of neurons in the first and second layers of the NN, learning ratio and slope of the activation function. This model addresses the challenge of optimizing a multi-output NN in a self-adjusted way. Performance of the developed network is evaluated by application to both univariate and multivariate traffic flow data from an urban highway. Results are analyzed based on the performance measures, showing that the genetic algorithm tunes the NN as well without any manually pre-adjustment. The achieved prediction accuracy is calculated with multiple measures such as the root mean square error (RMSE), and the RMSE value is 10 and 12 in the best configuration of the proposed model for single and multi-step-ahead traffic flow prediction, respectively. Document type: Article