Dynamic and Interpretable Hazard-Based Models of Traffic Incident Durations

Traffic incidents are a major contributor to congestion with as much as 54% of overall congestion to have been caused by traffic accidents or incidents in Florida, United States. In Brisbane, traffic incidents alone contribute to 50% of road congestion which is estimated to cost the community upwards of $400 million annually. There are more than 2500 traffic incidents recorded each year in the Brisbane region, half of which are minor or non-injury, which takes in excess of 2000 hrs to clear. Traffic incident management is the planning and coordination of people and material resources to safely and quickly address disruptions in the flow within the traffic network. The aim of this study is to research the factors that affect the traffic incident duration in the Port of Brisbane region, with the interest of heavy vehicle related crashes, incidents which obstructs traffic flow and incident durations longer than 90 minutes. Visits were made to Port of Brisbane to find the proportion of different types of vehicles and which of them has most significant impact on the study area. Most of the traffic incident data were collected from BMTMC. The final scope of this research was narrowed down to incident durations longer than 60 minutes that only involves heavy vehicle related crashes and incidents which obstruct traffic flow. Therefore out of 1834 incidents, only 27 of them were relevant to our topic. Factors affecting the duration of traffic incident management around the area of Port of Brisbane are response time, duration of site investigation and clearance time. Factors affecting the duration of traffic incident in Port of Brisbane cannot be determined due to the lack of data, however, from the obtained information, it can be seen that the verification and response time can be improved.

To date, studies of incidents on British motorways have been very limited. As part of a study into the role of Incident Support Units (ISUs) on the M25, the London orbital motorway, an analysis of road traffic incidents was undertaken. Motorway incident data were collected through closed circuit television (CCTV) observations at Surrey Police's Godstone Motorway Control Room. The collected data covered the Surrey Police section of the M25 motorway between Clacket Lane Services and Junction 14, for a four-week period between 5 November and 11 December 2003. Incident frequency, durations, locations, detection times, notification methods and ISU involvement were examined. It was found that the average incident frequency was 16·1 incidents per day for the study area, and the highest occurrence of incidents was found to be between 11·00 am and 12·00 pm. On average the recorded incidents lasted 1 h 9 min, with those occurring in the evening peak period lasting the longest. The distribution of incidents along the study area showed that the section between Junctions 9 and 10 had the highest rate for the study period. A comparison of incident characteristics between US and UK studies was also conducted. Incident rates between the M25 and the US were found to be substantially different, and incident durations were found to be between double and triple those of the US. Possible reasons for these differences are discussed in the text.

Prediction of incident duration is very important in Advanced Traffic Incident Management (ATIM). Predicting accurately the duration of a traffic incident is necessary to effectively reroute traffic around the incident and to clear traffic, in general, away from the incident's area. In this paper, Support Vector Regression (SVR) is employed to predict the incident duration. In the experiments, the authors used one incident data set collected from an expressway in the Netherlands. The data set is divided into two parts; one is for model development, and the other is for the model validation. The experiment results, including error analysis, indicate that the prediction model based on Support Vector Regression (SVR) can obtain high accuracy for incident duration. This model can effectively be applied to traffic incident detection and clearance systems.

Assessing and prioritizing the duration time and effects of traffic incidents on major roads present significant challenges for road network managers. This study examines the effect of numerous factors associated with various types of incidents on their duration and proposes an incident duration prediction model. Several parametric accelerated failure time hazard-based models were examined, including Weibull, log-logistic, log-normal, and generalized gamma, as well as all models with gamma heterogeneity and flexible parametric hazard-based models with freedom ranging from one to ten, by analyzing a traffic incident dataset obtained from the Incident Reporting and Dispatching System in Beijing in 2008. Results show that different factors significantly affect different incident time phases, whose best distributions were diverse. Given the best hazard-based models of each incident time phase, the prediction result can be reasonable for most incidents. The results of this study can aid traffic incident management agencies not only in implementing strategies that would reduce incident duration, and thus reduce congestion, secondary incidents, and the associated human and economic losses, but also in effectively predicting incident duration time.

Traffic congestion which has been a common problem faced by almost all of the metropolitan cities across the world is caused by non-recurring traffic incidents such as accidents and vehicle breakdowns in many cases. Therefore, accurate prediction of the incident duration has become an essential part of Intelligent Transport Systems (ITS). Although the data-driven deterministic regression models can serve this purpose, the performances of these models are subject to varying test conditions. In this study, we consider two different data-sets, one from Singapore and the other from the Netherlands, and explore stochastic models (Bayesian Support Vector Regression, i.e. BSVR and Gaussian Process, i.e. GP) that not only perform prediction but also provide error bars as confidence measures along with the predicted values. Moreover, we also verify whether these methods can anticipate when the predictions are unreliable. To this end, we analyze the specificity and sensitivity of the two methods for detecting prediction errors with different tolerance levels. We observe that for 70% specificity, the sensitivity of BSVR and GP is 67% and 78% respectively, for the incidents in Singapore. On the other hand, for the incidents in the Netherlands, the two methods have 55% and 60% sensitivity respectively, for the same level of specificity. Therefore, our proposed method of Bayesian models with error bars leads to more reliable predictions of the traffic incident duration, as well as forecasts to some extent when the predictions are unreliable.

Incidents impose substantial social and personal costs on drivers. Some of the larger incidents that cause delays are also associated with secondary incidents. However, the nature of interdependence between primary and secondary incidents is not fully known. The objective of this study is to understand how primary incident duration and secondary incident occurrence are related. Specifically, secondary incidents are more likely to occur if the primary incident lasts a long time; at the same time, the durations of primary incidents are expected to be longer if secondary incidents occur. After traffic incident and road inventory data in the Hampton Roads, Virginia, area were obtained, secondary incidents were then identified. Secondary incidents were defined as incidents occurring on the same roadway segment (which average 1 mi in length) as the primary incident and within the actual duration of the primary incident. If the primary incident blocked lanes, then the actual duration plus 15 min was used as the threshold. Models for primary incident durations and whether a secondary incident occurs are estimated. The interdependence is modeled by considering incident duration as endogenous in the secondary incident occurrence models. Results show statistical evidence for interdependence, but when it is taken into account, no substantial differences in the magnitudes and statistical significance for the estimated independent variables are found (compared with when the interdependence is not accounted for). Statistically significant correlations found between secondary incident occurrence and other variables allow the recommendation of specific operational strategies.

Prediction of incident duration is very important in Advanced Intelligent Traffic Incident Management and the accuracy of prediction can provide exact information for travellers. It is widely used in the area of ITS. In this paper, K-Nearest neighbor (KNN) is employed to predict the incident duration, which puts forward a new distance metric and weight determination method. This KNN model is created based on the incident data set collected by DVS-Center for Transport and Navigation, Ministry of Transport, Public Works and Management, the Netherlands. Moreover, a simulation based on Matlab is used for incident duration prediction and optimizing the best k value. Finally, an error analysis is made based on this simulation. As a result, this method (KNN) obtains high accuracy and has a better effect than Bayesian Decision Method-Based Tree Algorithm. So it can be effectively applied to intelligent traffic incident detection and clearance systems.

The prediction of traffic incident duration is an important foundation of advanced incident management system and driver information system. In this paper, actual traffic incident data was used to study the prediction problem of traffic incident duration by the method of neural network. 660 sets of actual traffic incident data from a freeway management center were used to train a neural network model, and 170 sets of incident data in the same data collection, which are different from training data, were used to test the prediction effect of the model. The test result shows that the correlation of the prediction values and the actual values is 0.8535, which indicates that the prediction result of the neural network model can basically represent actual incident duration.

In this paper, two types of incidents, vehicle breakdown and traffic crash, are studied. First, the characteristics of the incidents are analysed. Secondly, the potential factors that may affect incident duration such as vehicle type, incident time, incident location, number of vehicles involved, and incident severity, are discussed. Based on the analysis of the characteristics of the incidents, an incident duration model is developed. This model is based on fuzzy logic theory, which is a promising mathematical approach to model the process characterised by subjectivity, ambiguity, uncertainty, and imprecision, which can be seen in the incident duration. Analysis shows that the incident duration has a Weibull distribution. Although, there is no significant difference in duration between the vehicle breakdown and traffic crash, the factors that affect duration are different. The result shows that fuzzy logic models have good performance to predict the incident duration. It provides a different approachs to analyse the incident duration other than the statistical methods.

In modern power system, along with the developments of the data collecting technologies, the intensive and high-dimensional load data collection can be achieved. Therefore, to deeply reveal the patterns and behaviors hidden in the load dataset using load classification is of great significance for improving the service quality and the user experience of the power system. However, inevitable issues, for example the data missing and class imbalance are frequently reported in the present load dataset, which deteriorates the performance of the classification algorithms. Also, due to the special features, for example the time series, periodicity, and fluctuation of the load data, the traditional data classification algorithms also encounter performance defects. Therefore, this paper presents a data augmentation based enhanced temporal convolutional network (TCN) algorithm in enabling load classification. In the data augmentation phase, first an LRTC-TSVD algorithm is presented to implement the missing data completion. Second, a WGAN based class balancing approach is further presented to solve the class imbalance issue. Then, in the enhanced TCN phase, a WeightNorm, exponential linear unit (ELU) activation function, residual connection, and bidirectional feature fusion techniques based improved TCN (ITCN) algorithm is presented to carry out the accurate load data classification. Combining the data augmentation and the enhanced TCN phases, the ITCN algorithm is finally conducted. Based on the benchmark load datasets, the performances of the presented ITCN are evaluated. The experimental results report that the presented data augmentations can improve the quality of the dataset, moreover the classification algorithm is able to achieve the satisfied classification accuracy.

Air pollution is an environmental problem facing mankind today. Therefore, predicting the concentration of air pollutants in advance plays an important role in people's life and government decision-making. In this paper, a multi-channel asymmetric structure prediction model based on temporal convolutional network (TCN) is proposed. As TCN omits some feature information when learning time series features, increasing the number of channels will improve the receptive field of the model, cover longer historical information and extract more time series features. The influence of meteorological factors on the concentration of air pollutants is fully considered in the prediction model, which is used as an auxiliary factor to improve the prediction performance of the model. The concentration of air pollutants collected from the air monitoring station in Fushun City, Liaoning Province, is used as the data set to verify the effectiveness of the model, and the experimental comparison with other prediction models is conducted. The results show that the model proposed in this paper has more accurate prediction accuracy and stronger stability.

Traffic incidents are recognised as one of the key sources of non-recurrent congestion that often leads to reduction in travel time reliability (TTR), a key metric of roadway performance.A method is proposed here to quantify the impacts of traffic incidents on TTR on freeways.The method uses historical data to establish recurrent speed profiles and identifies nonrecurrent congestion based on their negative impacts on speeds.The locations and times of incidents are used to identify incidents among non-recurrent congestion events.Buffer time is employed to measure TTR.Extra buffer time is defined as the extra delay caused by traffic incidents.This reliability measure indicates how much extra travel time is required by travellers to arrive at their destination on time with 95% certainty in the case of an incident, over and above the travel time that would have been required under recurrent conditions.An extra buffer time index (EBTI) is defined as the ratio of extra buffer time to recurrent travel time, with zero being the best case (no delay).A Tobit model is used to identify and quantify factors that affect EBTI using a selected freeway segment in the Southeast Queensland, Australia network.Both fixed and random parameter Tobit specifications are tested.The estimation results reveal that models with random parameters offer a superior statistical fit for all types of incidents, suggesting the presence of unobserved heterogeneity across segments.What factors influence EBTI depends on the type of incident.In addition, changes in TTR as a result of traffic incidents are related to the characteristics of the incidents (multiple vehicles involved, incident duration, major incidents, etc.) and traffic characteristics.

In order to study the influence of different vehicles to incident delay, traffic simulation is applied to make the delay predictions. Three major parts, respectively the establishment of simulation platform and parameter calibration, survey and calculation of traffic data, and case study, are described in the paper. For the establishment of simulation platform and parameter calibration, at first, the local parameter and the global parameter, which need to be calibrated while making the delay simulation, are chosen. Then, the performance evaluation index and the performance evaluation target are proposed and the minimum simulation times are determined based on the statistical analysis. Finally, the flow of parameter calibration is designed, which first calibrates the local parameter and then calibrates the global parameter. For the survey and calculation of traffic data, at first, based on the toll data, the calculation method of the vehicle data, the evaluation index data and the incident delay data, are proposed. Then, these data are taken as the true values to calibrate the parameters and validate the simulation delay. At the same time, the incident duration is predicted. For case study, this paper takes six traffic incidents which took place in link 170002 of Nanjing-Nantong freeway (hereinafter as Ning-Tong freeway) in 2010 as sample data to calibrate the parameters and simulate the delay, and further compares the simulation delay with the true delay. The simulation results indicate that the calibrated model is consistent with the real link, the calculation of incident delay based on the simulation is more suitable and accurate, and this delay prediction is convenient to assist the freeway operators to make strategic decisions for the incident management.

Traffic incidents occur frequently on urban roadways and cause incident-induced congestion. Predicting incident duration is a key step in managing these events. Ordinary least squares (OLS) regression models can be estimated to relate the mean of incident duration data with its correlates. Because of the presence of larger incidents, duration distributions are often right-skewed; that is, the OLS model underpredicts the durations of larger incidents. Therefore, this study applies a modeling technique known as quantile regression to predict more accurately the skewed distribution of incident durations. Quantile regression estimates the relationships between correlates and a chosen percentile—for example, the 75th or 95th percentile—while the OLS regression is based on the mean of incident duration. With the use of incident data related to more than 85,000 (2013 to 2015) incidents for highways in the Hampton Roads area of Virginia, quantile regression results indicate that the magnitudes of parameters and predictions can be quite different compared with OLS regression. In addition to predicting durations of larger incidents more accurately, quantile regressions can estimate the probability of an incident lasting for a specific duration; for example, incidents involving congestion and delay have an approximately 25% chance of lasting more than 100.8 min, while incidents excluding congestion and delay are estimated to have a 25% chance of lasting more than 43.3 min. Such information is helpful in accurately predicting durations and developing potential applications for using quantile regressions for better traffic incident management.

The presence of anomalies in time series, i.e., data points that deviate from the expected behavior, can be a symptom of an incoming failure that may lead to costly repair. Thus, the detection of anomalous data by means of diagnostic methodologies can both reduce maintenance actions and asset unscheduled stops. To tackle this challenge, for the first time in the literature, we exploit the capabilities of Convolutional Neural Networks (CNNs), by feeding them with images obtained from multivariate time series data, transformed by means of two different approaches, namely, Gramian Angular Summation Field (GASF) and Markov Transition Field (MTF). Two CNN architectures are investigated, i.e., VGG-19 and SqueezeNet. The performance of both CNNs fed with images is compared to that of i) a Temporal Convolutional Network (TCN) fed with time series data and ii) a Support Vector Machine (SVM) model. In this paper, we present the comprehensive framework, which starts from time series transformation, goes through CNN development and ends with anomaly detection. The framework is applied to field data taken during normal operation of ten SGT-800 gas turbines, operated in combined cycle, running at base load, and located in two different regions. The normal data covers 150 days of operation. Spike faults are implanted in two out of the twenty available measured variables, i.e., compressor discharge temperature and compressor discharge pressure, by considering nine combinations of maximum fault magnitude and number of implanted spikes in each time series. The analyses carried out in this paper demonstrate that both CNNs fed with images achieve significantly higher classification accuracy than both a TCN model fed with time series data and an SVM model. Moreover, the MTF method always proves more robust than GASF method, and also allows higher accuracy values, in the range from 0.85 to 0.99.