More robust and better: a multiple kernel support vector machine ensemble approach for traffic incident detection

This paper presents a multiple kernel support vector machine (MKL‐SVM) ensemble algorithm to detect traffic incidents. It uses resampling technology to generate training set, test set, and training subset firstly; then uses different training subsets to train individual MKL‐SVM classifiers; and finally introduces ensemble methods to construct MKL‐SVM ensemble to detect traffic incidents. Extensive experiments have been performed to evaluate the performances of the four algorithms: standard SVM, SVM ensemble, MKL‐SVM, and the proposed algorithm (MKL‐SVM ensemble). The experimental results show that the proposed algorithm has the best comprehensive performances in traffic incidents detection. To achieve better performances, the proposed algorithm needs less individual classifiers to construct the ensemble than SVM ensemble algorithm. Thus, compared with SVM ensemble algorithm, the complexity of the ensemble classifier of the proposed algorithm is reduced greatly. Conveniently, the proposed algorithm also avoids the burden of selecting the appropriate kernel function and parameters. Copyright © 2013 John Wiley & Sons, Ltd.     

A computationally efficient two-stage method for short-term traffic prediction on urban roads

Abstract

Short-term traffic prediction plays an important role in intelligent transport systems. This paper presents a novel two-stage prediction structure using the technique of Singular Spectrum Analysis (SSA) as a data smoothing stage to improve the prediction accuracy. Moreover, a novel prediction method named Grey System Model (GM) is introduced to reduce the dependency on method training and parameter optimisation. To demonstrate the effects of these improvements, this paper compares the prediction accuracies of SSA and non-SSA model structures using both a GM and a more conventional Seasonal Auto-Regressive Integrated Moving Average (SARIMA) prediction model. These methods were calibrated and evaluated using traffic flow data from a corridor in Central London under both normal and incident traffic conditions. The prediction accuracy comparisons show that the SSA method as a data smoothing step before the application of machine learning or statistical prediction methods can improve the final traffic prediction accuracy. In addition, the results indicate that the relatively novel GM method outperforms SARIMA under both normal and incident traffic conditions on urban roads.     

Urban traffic flow prediction: a spatio‐temporal variable selection‐based approach

Short‐term traffic flow prediction in urban area remains a difficult yet important problem in intelligent transportation systems. Current spatio‐temporal‐based urban traffic flow prediction techniques trend aims to discover the relationship between adjacent upstream and downstream road segments using specific models, while in this paper, we advocate to exploit the spatial and temporal information from all available road segments in a partial road network. However, the available traffic states can be high dimensional for high‐density road networks. Therefore, we propose a spatio‐temporal variable selection‐based support vector regression (VS‐SVR) model fed with the high‐dimensional traffic data collected from all available road segments. Our prediction model can be presented as a two‐stage framework. In the first stage, we employ the multivariate adaptive regression splines model to select a set of predictors most related to the target one from the high‐dimensional spatio‐temporal variables, and different weights are assigned to the selected predictors. In the second stage, the kernel learning method, support vector regression, is trained on the weighted variables. The experimental results on the real‐world traffic volume collected from a sub‐area of Shanghai, China, demonstrate that the proposed spatio‐temporal VS‐SVR model outperforms the state‐of‐the‐art. Copyright © 2015 John Wiley & Sons, Ltd.     

Short‐term highway traffic flow prediction based on a hybrid strategy considering temporal–spatial information

Short‐term traffic flow prediction is fundamental for the intelligent transportation system and is proved to be a challenge. This paper proposed a hybrid strategy that is general and can make use of a large number of underlying machine learning or time‐series prediction models to capture the complex patterns beneath the traffic flow. With the strategy, four different combinations were implemented. To consider the spatial features of traffic phenomenon, several different state vectors including different observations were built. The performance of the proposed strategy was investigated using the traffic flow measurements from the Traffic Operation and Safety Laboratory in Wisconsin, USA. The results show the overall performance of hybrid strategy is better than a single model. Also, incorporating observations from adjacent junctions can improve prediction accuracy. Copyright © 2017 John Wiley & Sons, Ltd.     

Real‐time identification of traffic conditions prone to injury and non‐injury crashes on freeways using genetic programming

This study applied the genetic programming (GP) model to identify traffic conditions prone to injury and property‐damage‐only (PDO) crashes in different traffic states on freeways. It was found that the traffic conditions prone to injury and PDO crashes can be classified into a high‐speed and a low‐speed traffic state. The random forest (RF) analyses were conducted to identify the contributing factors to injury and PDO crashes in these two traffic states. Four separate GP models were then developed to link the risks of injury and PDO crashes in two traffic states to the variables selected by the RF. An overall GP model was also developed for the combined dataset. It was found that the separate GP models that considered different traffic states and crash severity provided better predictive performance than the overall model, and the traffic flow variables that affected injury and PDO crashes were quite different across different traffic states. The proposed GP models were also compared with the traditional logistic regression models. The results suggested that the GP models outperformed the logistic regression models in terms of the prediction accuracy. More specifically, the GP models increased the prediction accuracy of injury crashes by 10.7% and 8.0% in the low‐speed and high‐speed traffic states. For PDO crashes, the GP models increased the prediction accuracy by 7.4% and 6.0% in the low‐speed and high‐speed traffic states. Copyright © 2016 John Wiley & Sons, Ltd.     

Short‐term traffic flow prediction with linear conditional Gaussian Bayesian network

Traffic flow prediction is an essential part of intelligent transportation systems (ITS). Most of the previous traffic flow prediction work treated traffic flow as a time series process only, ignoring the spatial relationship from the upstream flows or the correlation with other traffic attributes like speed and density. In this paper, we utilize a linear conditional Gaussian (LCG) Bayesian network (BN) model to consider both spatial and temporal dimensions of traffic as well as speed information for short‐term traffic flow prediction. The LCG BN allows both continuous and discrete variables, which enables the consideration of categorical variables in traffic flow prediction. A microscopic traffic simulation dataset is used to test the performance of the proposed model compared to other popular approaches under different predicting time intervals. In addition, the authors investigate the importance of spatial data and speed data in flow prediction by comparing models with different levels of information. The results indicate that the prediction accuracy will increase significantly when both spatial data and speed data are included. Copyright © 2016 John Wiley & Sons, Ltd.     

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

Short-term traffic flow prediction is an integral part in most of Intelligent Transportation Systems (ITS) research and applications. Many researchers have already developed various methods that predict the future traffic condition from the historical database. Nevertheless, there has not been sufficient effort made to study how to identify and utilize the different factors that affect the traffic flow. In order to improve the performance of short-term traffic flow prediction, it is necessary to consider sufficient information related to the road section to be predicted. In this paper, we propose a method of constructing traffic state vectors by using mutual information (MI). First, the variables with different time delays are generated from the historical traffic time series, and the spatio-temporal correlations between the road sections in urban road network are evaluated by the MI. Then, the variables with the highest correlation related to the target traffic flow are selected by using a greedy search algorithm to construct the traffic state vector. The K-Nearest Neighbor (KNN) model is adapted for the application of the proposed state vector. Experimental results on real-world traffic data show that the proposed method of constructing traffic state vector provides good prediction accuracy in short-term traffic prediction.     

A hybrid deep learning based traffic flow prediction method and its understanding

Deep neural networks (DNNs) have recently demonstrated the capability to predict traffic flow with big data. While existing DNN models can provide better performance than shallow models, it is still an open issue of making full use of spatial-temporal characteristics of the traffic flow to improve their performance. In addition, our understanding of them on traffic data remains limited. This paper proposes a DNN based traffic flow prediction model (DNN-BTF) to improve the prediction accuracy. The DNN-BTF model makes full use of weekly/daily periodicity and spatial-temporal characteristics of traffic flow. Inspired by recent work in machine learning, an attention based model was introduced that automatically learns to determine the importance of past traffic flow. The convolutional neural network was also used to mine the spatial features and the recurrent neural network to mine the temporal features of traffic flow. We also showed through visualization how DNN-BTF model understands traffic flow data and presents a challenge to conventional thinking about neural networks in the transportation field that neural networks is purely a “black-box” model. Data from open-access database PeMS was used to validate the proposed DNN-BTF model on a long-term horizon prediction task. Experimental results demonstrated that our method outperforms the state-of-the-art approaches.     

Short‐term prediction of traffic parameters—performance comparison of a data‐driven and less‐data‐required approaches

The travel decisions made by road users are more affected by the traffic conditions when they travel than the current conditions. Thus, accurate prediction of traffic parameters for giving reliable information about the future state of traffic conditions is very important. Mainly, this is an essential component of many advanced traveller information systems coming under the intelligent transportation systems umbrella. In India, the automated traffic data collection is in the beginning stage, with many of the cities still struggling with database generation and processing, and hence, a less‐data‐demanding approach will be attractive for such applications, if it is not going to reduce the prediction accuracy to a great extent. The present study explores this area and tries to answer this question using automated data collected from field. A data‐driven technique, namely, artificial neural networks (ANN), which is shown to be a good tool for prediction problems, is taken as an example for data‐driven approach. Grey model, GM(1,1), which is also reported as a good prediction tool, is selected as the less‐data‐demanding approach. Volume, classified volume, average speed and classified speed at a particular location were selected for the prediction. The results showed comparable performance by both the methods. However, ANN required around seven times data compared with GM for comparable performance. Thus, considering the comparatively lesser input requirement of GM, it can be considered over ANN in situations where the historic database is limited. Copyright © 2016 John Wiley & Sons, Ltd.     

Adaptive traffic control system: Control strategy,prediction, resolution,and accuracy

Recent advances in traffic control methods have led to flexible control strategies for use in an adaptive traffic control system (ATCS). ATCS aims at controlling the imminent traffic, which is yet to arrive and hence not known perfectly. Therefore, volume prediction is an essential part. Associated with the prediction are two aspects: resolution and accuracy. Recent studies indicate a tradeoff between prediction resolution and accuracy: finer resolutions, larger errors. It is imperative to study the relationship and tradeoff between the control strategy, prediction resolution, and its associated error, which are crucial to the development of ATCS. This study investigates this relationship through an extensive simulation of scenarios in Hong Kong with a recently developed dynamic traffic control model, DISCO. Based on the Hong Kong scenarios conducted with DISCO, the major findings include: (i) the importance of resolution outweighs that of error; (ii) dynamic timing plans generally outperform time‐invariant timing plans; (iii) up to a certain extent, overestimated predictions lead to better results than underestimated predictions.     

Spatiotemporal traffic forecasting: review and proposed directions

This paper systematically reviews studies that forecast short-term traffic conditions using spatial dependence between links. We extract and synthesise 130 research papers, considering two perspectives: (1) methodological framework and (2) methods for capturing spatial information. Spatial information boosts the accuracy of prediction, particularly in congested traffic regimes and for longer horizons. Machine learning methods, which have attracted more attention in recent years, outperform the naïve statistical methods such as historical average and exponential smoothing. However, there is no guarantee of superiority when machine learning methods are compared with advanced statistical methods such as spatiotemporal autoregressive integrated moving average. As for the spatial dependency detection, a large gulf exists between the realistic spatial dependence of traffic links on a real network and the studied networks as follows: (1) studies capture spatial dependency of either adjacent or distant upstream and downstream links with the study link, (2) the spatially relevant links are selected either by prejudgment or by correlation-coefficient analysis, and (3) studies develop forecasting methods in a corridor test sample, where all links are connected sequentially together, assume a similarity between the behaviour of both parallel and adjacent links, and overlook the competitive nature of traffic links.     

Local online kernel ridge regression for forecasting of urban travel times

Accurate and reliable forecasting of traffic variables is one of the primary functions of Intelligent Transportation Systems. Reliable systems that are able to forecast traffic conditions accurately, multiple time steps into the future, are required for advanced traveller information systems. However, traffic forecasting is a difficult task because of the nonlinear and nonstationary properties of traffic series. Traditional linear models are incapable of modelling such properties, and typically perform poorly, particularly when conditions differ from the norm. Machine learning approaches such as artificial neural networks, nonparametric regression and kernel methods (KMs) have often been shown to outperform linear models in the literature. A bottleneck of the latter approach is that the information pertaining to all previous traffic states must be contained within the kernel, but the computational complexity of KMs usually scales cubically with the number of data points in the kernel. In this paper, a novel kernel-based machine learning (ML) algorithm is developed, namely the local online kernel ridge regression (LOKRR) model. Exploiting the observation that traffic data exhibits strong cyclic patterns characterised by rush hour traffic, LOKRR makes use of local kernels with varying parameters that are defined around each time point. This approach has 3 advantages over the standard single kernel approach: (1) It allows parameters to vary by time of day, capturing the time varying distribution of traffic data; (2) It allows smaller kernels to be defined that contain only the relevant traffic patterns, and; (3) It is online, allowing new traffic data to be incorporated as it arrives. The model is applied to the forecasting of travel times on London’s road network, and is found to outperform three benchmark models in forecasting up to 1 h ahead.     

Integrated model for traffic flow forecasting under rainy conditions

Inclement weather, such as heavy rain, significantly affects road traffic flow operation, which may cause severe congestion in road networks in cities. This study investigates the effect of inclement weather, such as rain events, on traffic flow and proposes an integrated model for traffic flow parameter forecasting during such events. First, an analysis of historical observation data indicates that the forecasting error of traffic flow volume has a significant linear correlation with mean precipitation, and thus, forecasting accuracy can be considerably improved by applying this linear correlation to correct forecasting values. An integrated online precipitation‐correction model was proposed for traffic flow volume forecasting based on these findings. We preprocessed precipitation data transformation and used outlier detection techniques to improve the efficiency of the model. Finally, an integrated forecasting model was designed through data fusion methods based on the four basic forecasting models and the proposed online precipitation‐correction model. Results of the model validation with the field data set show that the designed model is better than the other models in terms of overall accuracy throughout the day and under precipitation. However, the designed model is not always ideal under heavy rain conditions. Copyright © 2016 John Wiley & Sons, Ltd.     

Efficient missing data imputing for traffic flow by considering temporal and spatial dependence

The missing data problem remains as a difficulty in a diverse variety of transportation applications, e.g. traffic flow prediction and traffic pattern recognition. To solve this problem, numerous algorithms had been proposed in the last decade to impute the missed data. However, few existing studies had fully used the traffic flow information of neighboring detecting points to improve imputing performance. In this paper, probabilistic principle component analysis (PPCA) based imputing method, which had been proven to be one of the most effective imputing methods without using temporal or spatial dependence, is extended to utilize the information of multiple points. We systematically examine the potential benefits of multi-point data fusion and study the possible influence of measurement time lags. Tests indicate that the hidden temporal–spatial dependence is nonlinear and could be better retrieved by kernel probabilistic principle component analysis (KPPCA) based method rather than PPCA method. Comparison proves that imputing errors can be notably reduced, if temporal–spatial dependence has been appropriately considered.     

A new probability statistical model for traffic noise prediction on free flow roads and control flow roads

A new traffic noise prediction approach based on a probability distribution model of vehicle noise emissions and achieved by Monte Carlo simulation is proposed in this paper. The probability distributions of the noise emissions of three types of vehicles are obtained using an experimental method. On this basis, a new probability statistical model for traffic noise prediction on free flow roads and control flow roads is established. The accuracy of the probability statistical model is verified by means of a comparison with the measured data, which has shown that the calculated results of L_eq, L₁₀, L₅₀, L₉₀, and the probability distribution of noise level occurrence agree well with the measurements. The results demonstrate that the new method can avoid the complicated process of traffic flow simulation but still maintain high accuracy for the traffic noise prediction.     

A novel variable selection method based on frequent pattern tree for real-time traffic accident risk prediction

With the availability of large volumes of real-time traffic flow data along with traffic accident information, there is a renewed interest in the development of models for the real-time prediction of traffic accident risk. One challenge, however, is that the available data are usually complex, noisy, and even misleading. This raises the question of how to select the most important explanatory variables to achieve an acceptable level of accuracy for real-time traffic accident risk prediction. To address this, the present paper proposes a novel Frequent Pattern tree (FP tree) based variable selection method. The method works by first identifying all the frequent patterns in the traffic accident dataset. Next, for each frequent pattern, we introduce a new metric, herein referred to as the Relative Object Purity Ratio (ROPR). The ROPR is then used to calculate the importance score of each explanatory variable which in turn can be used for ranking and selecting the variables that contribute most to explaining the accident patterns. To demonstrate the advantages of the proposed variable selection method, the study develops two traffic accident risk prediction models, based on accident data collected on interstate highway I-64 in Virginia, namely a k-nearest neighbor model and a Bayesian network. Prior to model development, two variable selection methods are utilized: (1) the FP tree based method proposed in this paper; and (2) the random forest method, a widely used variable selection method, which is used as the base case for comparison. The results show that the FP tree based accident risk prediction models perform better than the random forest based models, regardless of the type of prediction models (i.e. k-nearest neighbor or Bayesian network), the settings of their parameters, and the types of datasets used for model training and testing. The best model found is a FP tree based Bayesian network model that can predict 61.11% of accidents while having a false alarm rate of 38.16%. These results compare very favorably with other accident prediction models reported in the literature.     

A network‐based dynamic air traffic flow model for short‐term en route traffic prediction

This paper presents a dynamic network‐based approach for short‐term air traffic flow prediction in en route airspace. A dynamic network characterizing both the topological structure of airspace and the dynamics of air traffic flow is developed, based on which the continuity equation in fluid mechanics is adopted to describe the continuous behaviour of the en route traffic. Building on the network‐based continuity equation, the space division concept in cell transmission model is introduced to discretize the proposed model both in space and time. The model parameters are sequentially updated based on the statistical properties of the recent radar data and the new predicting results. The proposed method is applied to a real data set from Shanghai Area Control Center for the short‐term air traffic flow prediction both at flight path and en route sector level. The analysis of the case study shows that the developed method can characterize well the dynamics of the en route traffic flow, thereby providing satisfactory prediction results with appropriate uncertainty limits. The mean relative prediction errors are less than 0.10 and 0.14, and the absolute errors fall in the range of 0 to 1 and 0 to 3 in more than 95% time intervals respectively, for the flight path and en route sector level. Copyright © 2017 John Wiley & Sons, Ltd.     

Predicting and planning airport acceptance rates in metroplex systems for improved traffic flow management decision support

Efficient planning of Airport Acceptance Rates (AARs) is key for the overall efficiency of Traffic Management Initiatives such as Ground Delay Programs (GDPs). Yet, precisely estimating future flow rates is a challenge for traffic managers during daily operations as capacity depends on a number of factors/decisions with very dynamic and uncertain profiles. This paper presents a data-driven framework for AAR prediction and planning towards improved traffic flow management decision support. A unique feature of this framework is to account for operational interdependency aspects that exist in metroplex systems and affect throughput performance. Gaussian Process regression is used to create an airport capacity prediction model capable of translating weather and metroplex configuration forecasts into probabilistic arrival capacity forecasts for strategic time horizons. To process the capacity forecasts and assist the design of traffic flow management strategies, an optimization model for capacity allocation is developed. The proposed models are found to outperform currently used methods in predicting throughput performance at the New York airports. Moreover, when used to prescribe optimal AARs in GDPs, an overall delay reduction of up to 9.7% is achieved. The results also reveal that incorporating robustness in the design of the traffic flow management plan can contribute to decrease delay costs while increasing predictability.     

Classification modeling approach for vehicle dynamics model‐integrated traffic simulation assessing surrogate safety

To assess safety impacts of untried traffic control strategies, an earlier study developed a vehicle dynamics model‐integrated (i.e., VISSIM‐CarSim‐SSAM) simulation approach and evaluated its performance using surrogate safety measures. Although the study found that the integrated simulation approach was a superior alternative to existing approaches in assessing surrogate safety, the computation time required for the implementation of the integrated simulation approach prevents it from using it in practice. Thus, this study developed and evaluated two types of models that could replace the integrated simulation approach with much faster computation time, feasible for real‐time implementation. The two models are as follows: (i) a statistical model (i.e., logit model) and (ii) a nonparametric approach (i.e., artificial neural network). The logit model and the neural network model were developed and trained on the basis of three simulation data sets obtained from the VISSIM‐CarSim‐SSAM integrated simulation approach, and their performances were compared in terms of the prediction accuracy. These two models were evaluated using six new simulation data sets. The results indicated that the neural network approach showing 97.7% prediction accuracy was superior to the logit model with 85.9% prediction accuracy. In addition, the correlation analysis results between the traffic conflicts obtained from the neural network approach and the actual traffic crash data collected in the field indicated a statistically significant relationship (i.e., 0.68 correlation coefficient) between them. This correlation strength is higher than that of the VISSIM only (i.e., the state of practice) simulation approach. The study results indicated that the neural network approach is not only a time‐efficient way to implementing the VISSIM‐CarSim‐SSAM integrated simulation but also a superior alternative in assessing surrogate safety. Copyright © 2014 John Wiley & Sons, Ltd.     

Improved waveform‐feature‐based vehicle classification using a single‐point magnetic sensor

Vehicle classification systems have important roles in applications related to real‐time traffic management. They also provide essential data and necessary information for traffic planning, pavement design, and maintenance. Among various classification techniques, the length‐based classification technique is widely used at present. However, the undesirable speed estimates provided by conventional data aggregation make it impossible to collect reliable length data from a single‐point sensor during real‐time operations. In this paper, an innovative approach of vehicle classification will be proposed, which achieved very satisfactory results on a single‐point sensor. This method has two essential parts. The first concerns with the procedure of smart feature extraction and selection according to the proposed filter–filter–wrapper model. The model of filter–filter–wrapper is adopted to make an evaluation on the extracted feature subsets. Meanwhile, the model will determine a nonredundant feature subset, which can make a complete reflection on the differences of various types of vehicles. In the second part, an algorithm for vehicle classification according to the theoretical basis of clustering support vector machines (C‐SVMs) was established with the selected optimal feature subset. The paper also uses particle swarm optimization (PSO), with the purpose of searching for an optimal kernel parameter and the slack penalty parameter in C‐SVMs. A total of 460 samples were tested through cross validation, and the result turned out that the classification accuracy was over 99%. In summary, the test results demonstrated that our vehicle classification method could enhance the efficiency of machine‐learning‐based data mining and the accuracy of vehicle classification. Copyright © 2014 John Wiley & Sons, Ltd.