Titan

—Due to the recent vast availability of transportation traffic data, major research efforts have been devoted to traffic prediction, which is useful in many applications such as urban planning, traffic management and navigations systems. Current prediction methods that independently train a model per traffic sensor cannot accurately predict traffic in every situation (e.g., rush hours, constructions and accidents) because there may not exist sufficient training samples per sensor for all situations. To address this shortcoming, our core idea is to explore the com-monalities of prediction tasks across multiple sensors who behave similarly in a specific traffic situation. Instead of building a model independently per sensor, we propose a Multi-Task Learning (MTL) framework that aims to first automatically identify the traffic situations and then simultaneously build one forecasting model for similar-behaving sensors per traffic situation. The key innovation here is that instead of the straightforward application of MTL where each " task " corresponds to a sensor, we relate each MTL's " task " to a traffic situation. Specifically, we first identify these traffic situations by running clustering algorithms on all sensors' data. Subsequently, to enforce the commonalities under each identified situation, we use the group Lasso regularization in MTL to select a common set of features for the prediction tasks, and we adapt efficient FISTA algorithm with guaranteed convergence rate. We evaluated our methods with a large volume of real-world traffic sensor data; our results show that by incorporating traffic situations, our proposed MTL framework performs consistently better than naively applying MTL per sensor. Moreover, our holistic approach, under different traffic situations, outperforms all the best traffic prediction approaches for a given situation by up to 18% and 30% in short and long term predictions, respectively.

2016

This paper describes joint work done by IBM Research (development of the solution) and GrandLyon (assessment of the solution) for real-time incident detection and traffic prediction on an urban road network in the presence of incidents. The detection and prediction methods are integrated and tested in a pilot study on live real-time traffic data in Lyon Metropolis, France. Numerical results and analysis are provided. The benefits of faster and more accurate incident detection include faster incident clearing times and more accurate and timely information to the public so as to avoid affected areas on the road network. 1.

World Journal of Advanced Research and Reviews, 2024

Accurate prediction of accident risks plays a crucial role in proactively implementing safety measures and allocating resources effectively. This paper introduces an innovative approach aimed at improving accident risk prediction by harnessing unique data sources and extracting insights from diverse yet sparse datasets. Traditional models often face limitations due to a lack of diversity and scope in the available data, which hinders their predictive capabilities. In response to this challenge, our study integrates a broad spectrum of heterogeneous data, encompassing traffic flow, weather conditions, road infrastructure details, and historical accident records. To overcome the difficulties associated with sparse data, we employ advanced data science techniques such as feature engineering, imputation, and machine learning. The paper introduces a novel dataset that amalgamates various data types, establishing a robust foundation for our predictive model. Through meticulous analysis, we derive valuable insights from these diverse sources, significantly enhancing our ability to assess accident risks. The proposed approach offers numerous advantages, including the capacity to predict accidents in areas that were previously underrepresented and under varying conditions. We rigorously evaluate the model's performance through extensive experimentation and validate its accuracy using real-world accident data. Our results indicate substantial improvements in prediction accuracy compared to conventional models. This research contributes significantly to the field of accident risk prediction by highlighting the potential benefits of integrating heterogeneous sparse data and leveraging advanced data science techniques. The study underscores the importance of tapping into novel data sources and extracting concealed patterns and insights to promote safety and optimize resource allocation in accident-prone regions, fostering more secure environments.

Iraqi Journal of Computer, Communication, Control and System Engineering, 2021

Traffic incidents dont only cause various levels of traffic congestion but often contribute to traffic accidents and secondary accidents, resulting in substantial loss of life, economy, and productivity loss in terms of injuries and deaths, increased travel times and delays, and excessive consumption of energy and air pollution. Therefore, it is essential to accurately estimate the duration of the incident to mitigate these effects. Traffic management center incident logs and traffic sensors data from Eastbound Interstate 70 (I-70) in Missouri, United States collected during the period from January 2015 to January 2017, with a total of 352 incident records were used to develop incident duration estimation models. This paper investigated different machine learning (ML) methods for traffic incidents duration prediction. The attempted ML techniques include Support Vector Machine (SVM), Random Forest (RF), and Neural Network Multi-Layer Perceptron (MLP). Root mean squared error (RMSE) a...

—The time wasted in traffic congestion especially the non-recurring (unexpected) is priceless. Intelligent Transportation Systems (ITS) technologies aim at improving transportation safety and mobility, reduce environmental impacts, and enhance productivity by utilizing the current available infrastructure. Providing the drivers with an accurate clearance time will help them manage their trips efficiently. Existing studies attempted to predict the clearance time of highway incidents through statistical modeling by fitting the incident features to their models. Unfortunately , the incident features are not available at the beginning of the incident. In this study, we develop efficient prediction models by training millions of real-time traffic and weather data from Florida Department of Transportation District 4 (FDOT-D4), in conjunction with the crash data reported for the same period. Our models can predict the incident clearance time in real-time once the accident happens by using the real-time traffic data. In this paper, we mainly use Deep Leaning (DL) to build the clearance time prediction model. In addition, we build two other models, Distributed Random Forests (DRF)and Generalized Linear Model (GLM) then, we compare the results of all these models. Traffic Management Centers (TMCs) will use the predicted results to display them on Dynamic Message Signs (DMSs) and broadcast them on the Highway Advisory Radio (HAR). The models performances are validated with the actual clearance times provided by the incidents respondents. DL outperforms both DRF and GLM in predicting the clearance time, reflected in the lowest root mean squared error (RMSE) of 0.99 and the lowest mean absolute error (MAE) of 0.6.

International Journal of Crashworthiness, 2019

With the exponential growth of the number of vehicles in the traffic, road traffic accidents have become a fast-growing and wide-spreading menace, causing the loss of precious human life and economic assets. In addition to the fast growth of population and motorization, roads are highly occupied by cars, buses, trucks, motorcycles, minibikes, taxis, pedestrians, animals, and other travelers. The main challenges in the road traffic accident data prediction and analysis is the small size of the dataset that can be used for training. While road traffic accident causes millions of deaths and injuries every year, their density in time and space is-fortunately-low. This makes machine learning very challenging at the local level. The main target of this work is to minimize the Causality Severity of road traffic accidents. Mathematical analysis for the attributes' effect on Causality Severity was guided to select effective twenty decision variables. An optimization search requires a reliable simulated model that depicts the necessary characteristics of the system under study. The good formulation of the simulated model was reduced the non-linearity and interaction between the variables. The simulated model has been verified by successive linear technique with aid of the Successive Linear Programming (SLP). Mixed Integer nonlinear Program (MINLP) has proven a reliable optimization search with the present model. Reasonable agreements have found when compared the simulated solutions with the experimental data. From the output results, we observed that some of the attributes have a positive effect on Causality-Severity, while other attributes have a negative effect on it. In addition to the implementation output of both MINLP and SLP, we can state that the Age of Driver, Speed Limit, and Age of Causality are necessary and sensitive variables for objective Causality Severity. Optimal results would guide for selecting the best conditions.

2007

Many deployed traffic incident detection systems use algorithms that require significant manual tuning. We seek machine learning incident detection solutions that reduce the need for manual adjustments by taking advantage of massive databases of traffic sensor network measurements. First, we show that a rather straightforward supervised learner based on the SVM model outperforms a fixed detection model used by state-of-the-art traffic incident detectors.

2020

Abstract: Predicting the traffic condition in urban networks is a priority for all traffic management centers around the world. This becomes very challenging especially when the network is affected by traffic incidents which vary in both time and space. Although data-driven machine learning (ML) modeling can be considered as an ideal tool for short-term traffic predictions, its performance is severely degraded when little historical traffic information is available under non-recurrent incident conditions. This paper addresses this challenge by integrating both data-driven and traffic simulation modeling. Instead of directly predicting the traffic states using limited historical data, we apply data-driven models to reinforce the traffic microsimulation. More explicitly, we employ ML models to predict origin-destination (OD) demand flows based on historical day-to-day demand flows. The traffic simulation uses the freshly reported incident information and the predicted OD demand flows ...

2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 2018

With the rapid growth in urban transit networks in recent years, detecting service disruptions in a timely manner is a problem of increased interest to service providers. Transit agencies are seeking to move beyond traditional customer questionnaires and manual service inspections to leveraging open source indicators like social media for deteting emerging transit events. In this paper, we leverage Twitter data for early detection of metro service disruptions. Inspired by the multi-task learning framework, we propose the Metro Disruption Detection Model, which captures the semantic similarity between transit lines in Twitter space. We propose novel constraints on feature semantic similarity exploiting prior knowledge about the spatial connectivity and shared tracks of the metro network. An algorithm based on the alternating direction method of multipliers (ADMM) framework is developed to solve the proposed model. We run extensive experiments and comparisons to other models with real world Twitter data and transit disruption records from the Washington Metropolitan Area Transit Authority (WMATA) to justify the efficacy of our model.

Journal of Big Data Analytics in Transportation

To discover the spatial and temporal traffic patterns, this paper proposes a spatiotemporal graphical modeling approach, spatiotemporal pattern network (STPN), to explore traffic dynamics in large traffic networks. A measurement based on Granger causality is used to identify the characteristics of spatial and temporal traffic patterns. An anomaly score is estimated to detect and locate traffic incidents in diverse types and severities, and also to quantify the influence of incidents on traffic flow fluctuations. Built upon symbolic dynamics filtering, the proposed approach implements spatial and temporal feature extraction via discovering causal dependencies among road segments using STPN, system-wide pattern learning through an energy-based model, restricted Boltzmann machine, and inference using a newly developed root-cause analysis algorithm. Case studies are carried out using the probe vehicle data collected on Interstate Highway 80 in Iowa and the results show that the proposed approach is capable of (1) discovering and representing causal interactions among subsystems (e.g., road segment) of a traffic network that provide valuable information for developing and applying customized traffic management strategies, (2) adaptively handling multiple nominal patterns mixed with anomalous data for effectively differentiating abnormal traffic system status and locating traffic incidents, and (3) quantifying the fluctuation of traffic flow and the severity of the detected incident via anomaly scores estimated from traffic speed behaviors. The findings from the case studies reiterate the importance of incorporating both temporal and spatial features for pattern analysis and incident detection. The proposed approach is built for real-time application and can be utilized for on-line incident detection.