Short-term traffic volume prediction by ensemble learning in concept drifting environments

As the fundamental part of other Intelligent Transportation Systems (ITS) applications, short-term traffic volume prediction plays an important role in various intelligent transportation tasks, such as traffic management, traffic signal control and route planning. Although Neural-network-based traffic prediction methods can produce good results, most of the models can’t be explained in an intuitive way. In this paper, we not only proposed a model that increase the short-term prediction accuracy of the traffic volume, but also improved the interpretability of the model by analyzing the internal attention score learnt by the model. we propose a spatiotemporal attention mechanism-based multistep traffic volume prediction model (SAMM). Inside the model, an LSTM-based Encoder-Decoder network with a hybrid attention mechanism is introduced, which consists of spatial attention and temporal attention. In the first level, the local and global spatial attention mechanisms considering the micro traffic evolution and macro pattern similarity, respectively, are applied to capture and amplify the features from the highly correlated entrance stations. In the second level, a temporal attention mechanism is employed to amplify the features from the time steps captured as contributing more to the future exit volume. Considering the time-dependent characteristics and the continuity of the recent evolutionary traffic volume trend, the timestamp features and historical exit volume series of target stations are included as the external inputs. An experiment is conducted using data from the highway toll collection system of Guangdong Province, China. By extracting and analyzing the weights of the spatial and temporal attention layers, the contributions of the intermediate parameters are revealed and explained with knowledge acquired by historical statistics. The results show that the proposed model outperforms the state-of-the-art model by 29.51% in terms of MSE, 13.93% in terms of MAE, and 5.69% in terms of MAPE. The effectiveness of the Encoder-Decoder framework and the attention mechanism are also verified.

This paper, based on the Kalman filter theory, established short-term traffic volume model. Kalman filter model has a good static stability, and adopts iterative method for optimal estimation of traffic. But the regular Kalman filtering traffic volume prediction established without considering the influencing factors on traffic and time-lag. Analyzing the influence factors of traffic volume based on grey entropy, and selects the main influencing factors by the size of grey entropy to establish the prediction of short-term traffic volume. Based on this, utilize the internet of things for data collection, and make a simulation experiment on a road in Shenyang. Simulation results show that model has good adaptability, high prediction accuracy on various states of traffic volume. It is a kind of effective traffic flow forecasting model.

Ferry traffic prediction is an important consideration for urban transportation authorities. A novel methodology which considers cyclical patterns in ferry on‐board vehicle traffic and improves on this method by incorporating traffic data from nearby freeway loop detectors is proposed. Since vehicles arriving in ferry terminals often originate from nearby freeways, the study of freeway traffic flow can improve ferry traffic prediction, especially during peak travel periods. In this study, the authors first identify periodic patterns in ferry traffic through analysis in both the time and frequency domain, and propose a modelling methodology which considers these patterns in short‐term traffic prediction. In the proposed model, the ferry traffic volume is divided into periodic and dynamic components, with the periodic component approximated by trigonometric functions. Second, this model is expanded to incorporate measured traffic data on nearby freeways. Specifically, the correlation between freeway traffic and ferry traffic volume is studied, and an enhanced model is proposed which integrates the predictions obtained through historical ferry data and from freeway traffic during peak travel time periods. Validation results demonstrate that consideration of periodic patterns increases prediction accuracy and that the proposed model further improves the ferry traffic prediction accuracy during peak time periods.

Traffic congestion at an intersection could be caused by the volume of vehicles that exceed the road capacity, the duration of the green light that is fixed, and so on. The volume of vehicles per unit time at an intersection cannot be known with certainty. Therefore, we need to predict it using fuzzy logic, specifically the Mamdani fuzzy implications. The problems are as follows: how are the input variables to be analyzed with Mamdani fuzzy implications; how are the prediction results, and how is the accuracy based on MAPE. The case study was conducted at two intersections in Semarang City. The tests were carried out using Matlab and manual calculations. The input variables in traffic volume prediction are MC, LV, HV, and time. While the input in the prediction of the duration of the green light is the number of motorcycles and cars. Based on the predictions, there are 74 vehicles (per hour) at the Kaligarang intersection in the east-north direction, there are 111 vehicles at the Kariadi intersection in the south-north direction, and the predictions are good and very accurate (measured by MAPE). The duration of the green light at the Kaligarang intersection on the west approach is 86 seconds, the duration on the Kariadi intersection on the north approach is 81 seconds, and the predictions are good and very accurate.

Intelligent transportation systems(ITS) is effective on solving the problem of traffic jam in cities. Prdiction of crossroads' traffic volume is the key technology in ITS. BP neural network is universally used in prediction of crossroads' traffic volume. This research aimed at using double-layers BP neural network to predict the traffic volume of Lishan Crossroad Jinan City. Results of the computer simulation showed that the method was applicable, the average relative tolerance was 9.71%. The doble-layers BP neural network can be used for prediction of crossroads' traffic volume.

Incremental learning with concept drift has often been tackled by ensemble methods, where models built in the past can be retrained to attain new models for the current data. Two design questions need to be addressed in developing ensemble methods for incremental learning with concept drift, i.e., which historical (i.e., previously trained) models should be preserved and how to utilize them. A novel ensemble learning method, namely, Diversity and Transfer-based Ensemble Learning (DTEL), is proposed in this paper. Given newly arrived data, DTEL uses each preserved historical model as an initial model and further trains it with the new data via transfer learning. Furthermore, DTEL preserves a diverse set of historical models, rather than a set of historical models that are merely accurate in terms of classification accuracy. Empirical studies on 15 synthetic data streams and 5 real-world data streams (all with concept drifts) demonstrate that DTEL can handle concept drift more effectively than 4 other state-of-the-art methods.

Prediction of traffic volume is the key technology in intelligent transportation systems. BP neural network is universally used in prediction of traffic volume. This research aimed at advancing BP neural networkpsilas precision in prediction of traffic flow. The method of prediction of traffic volume was based on the subsection learning of double-layers BP neural network. The improved method was used to predict the traffic volume of Jingshi road Jinan city, then compared the results maked by subsection-learning method and by common method. Using subsection-learning method,the average relative tolerance was decreased by 2.52%. The improved BP neural network can be used to predict the traffic volume.

Predictions for short-term traffic volume provide important inputs for traveler information and traffic management. Traffic volumes in the near future are often estimated based on historical volumes. Because of the complicated nonlinear relationship between historical and future traffic volume data, many previous studies used neural networks to predict short-term traffic volumes. In this research, a v-support vector machine (v-SVM) model, which has the particular strength of overcoming local minima and overfitting common to neural network models, is proposed for short-term traffic volume prediction. The v-SVM model is compared with a widely used multilayer feed-forward neural network (MLFNN) model using four data sets collected from three interstate freeways. Testing results show that for both one-step and two-step forecasting, the v-SVM model outperforms the MLFNN model for all data sets in terms of mean absolute percentage error and root-mean-square error. Key issues in applying both models are also discussed in this article.

Construction of traffic state vector using mutual information for short-term traffic flow prediction

A short-term traffic volume prediction model is proposed in this study. The model employs a multi-dimensional autoregressive type analysis so that not only the time series of traffic volume data on the study link of the road network but also traffic volume data on other links which may feed into the study link can be incorporated in the model. The similarity of the traffic flow pattern from day to day is taken into account in making up the prediction model. That is, the parameter vector h involved in the model to predict the traffic flow at k time intervals ahead of time is assumed smoothly changing with respect to day and identified by making use of kalman filtering theory.The proposed model is tested and compared against the two existing models, i.e., one utilizing the Box-Jenkins method and another based on spectral analysis, by using data collected from a network in Nagano City. The data are 15-min vehicle counts recorded at intervals of five minutes, the time unit in the models. The test results suggest that the proposed model performs substantially (up to 48% in terms of root mean square error of predictions) better than the existing models except in 5-min prediction. The tests also indicate a robustness in model performance as the prediction horizon increases, thus suggesting desirable behavior for longer term prediction.

Machine Learning Approach on Traffic Congestion Monitoring System in Internet of Vehicles

This paper proposes a unified spatio-temporal model on the basis of STARIMA (Space-Time AutoRegressive Integrated Moving Average) for short-term road traffic prediction. The contributions of this paper are as follows. First, we develop a physically intuitive approach to traffic prediction that captures the time-varying spatio-temporal correlation between traffic at different measurement points. The spatio-temporal correlation is affected by the road network topology, time-varying speed, and time-varying trip distribution. Distinctly different from previous black-box approaches to road traffic modeling and prediction, parameters of the proposed approach have physically intuitive meanings which make them readily amenable to suit changing road and traffic conditions. Second, unlike some existing techniques which capture the variation of spatio-temporal correlation by a complete redesign and calibration of the model, the proposed approach uses a unified model which incorporates the physical factors potentially affecting the variation of spatio-temporal correlation into a series of parameters. These parameters are relatively easy to control and adjust when road and traffic conditions change, thereby greatly reducing the computational complexity. Experiments using two set of real traffic traces demonstrate that the proposed approach has superior accuracy compared with the widely used ARIMA (AutoRegressive Integrated Moving Average) and is only marginally inferior to that obtained by constructing multiple STARIMA models for different time of the day, however with a much reduced computational and implementation complexity.

Generally, the planning of highway is designed on the basis of the traffic volume prediction. Because the influence factors of the traffic volume prediction are indeterminate, it leads to great discrepancy between the traffic prediction and the actual volume. Grey-markov forecasting model was founded by applying the model of GM (1,1) and Markov random process theory. The model utilizes the advantages of Grey-markov GM (1,1) forecasting model and Markov random process in order to discover the developing and varying tendency of the forecasting data sequences. When Markov random process is being used to decide the transfer rule of the state, it can not only make full use of the information of the historical data, but also improve the forecasting accuracy of random series. So it develops the applying range of grey forecasting and presents a new method of data sequences with large random. The analysis of an example indicates that the grey-markov model has good forecasting accuracy and excellent applicability.

A temporal data mining method was used to develop a short-term urban traffic flow volume prediction model in a signalized controlling situation, where the next green light phase traffic flow volume of a crossing entrance is predicted by historical green light phase traffic flow volume data of itself and upstream crossing entrances. An experiment is given, and the result is satisfying.KeywordsUrban traffic managementShort-term predictionData mining

Transportation projects require an enormous amount of capital investment resulting from their tremendous size, complexity, and risk. Due to the limitation of public finances, the private sector is invited to participate in transportation project development. The private sector can entirely or partially invest in transportation projects in the form of Public-Private Partnership (PPP) scheme, which has been an attractive option for several developing countries, including Vietnam. There are many factors affecting the success of PPP projects. The accurate prediction of traffic volume is considered one of the key success factors of PPP transportation projects. However, only few research works investigated how to predict traffic volume over a long period of time. Moreover, conventional traffic volume forecasting methods are usually based on deterministic models which predict a single value of traffic volume but do not consider risk and uncertainty. This knowledge gap makes it difficult for concessionaires to estimate PPP transportation project revenues accurately. The objective of this paper is to develop a probabilistic traffic volume prediction model. First, traffic volumes were estimated following the Geometric Brownian Motion (GBM) process. Monte Carlo technique is then applied to simulate different scenarios. The results show that this stochastic approach can systematically analyze variations in the traffic volume and yield more reliable estimates for PPP projects.