Enhanced delay propagation tree model with Bayesian Network for modelling flight delay propagation

An enhanced Delay Propagation Tree model with Bayesian Network (DPT-BN) is developed to model multi-flight delay propagation and delay interdependencies. Using a set of real airline data, results show that flights have non-homogeneous delay propagation effects. The DPT-BN model is used to infer posterior delay profiles with different delay and scheduling scenarios. The major contribution of the DPT-BN model is to demonstrate how the modelling of non-independent and identically distributed delay profiles is more realistic for the observed delay propagation mechanism, and how robust airline scheduling methodologies can benefit from this probability-based delay model.     

A Bayesian Network model for contextual versus non-contextual driving behavior assessment

Driving behavior is generally considered to be one of the most important factors in crash occurrence. This paper aims to evaluate the benefits of utilizing context-relevant information in the driving behavior assessment process (i.e. contextual driving behavior assessment approach). We use a Bayesian Network (BN) model that investigates the relationships between GPS driving observations, individual driving behavior, individual driving risks, and individual crash frequency. In contrast to prior studies without context information (i.e. non-contextual approach), the data used in the BN approach is a combination of contextual features in the surrounding environment that may contribute to crash risk, such as road conditions surrounding the vehicle of interest and dynamic traffic flow information, as well as the non-contextual data such as instantaneous driving speed and the acceleration/deceleration of a vehicle. An information-aggregation mechanism is developed to aggregates massive amounts of vehicle GPS data points, kinematic events and context information into drivel-level data. With the proposed model, driving behavior risks for drivers is assessed and the relationship between contextual driving behavior and crash occurrence is established. The analysis results in the case study section show that the contextual model has significantly better performance than the non-contextual model, and that drivers who drive at a speed faster than others or much slower than the speed limit at the ramp, and with more rapid acceleration or deceleration on freeways are more likely to be involved in crash events. In addition, younger drivers, and female drivers with higher VMT are found to have higher crash risk.     

Semi-autonomous vehicle motor insurance: A Bayesian Network risk transfer approach

All developed economies mandate at least third party auto insurance resulting inW a vast global liability industry. The evolution towards semi-autonomous and eventually driverless vehicles will progressively remove the leading cause of vehicle accidents, human error, and significantly lower vehicle accident rates. However, this transition will force a departure from existing actuarial methods requires careful management to ensure risks are correctly assigned. Personal motor insurance lines are anticipated to diminish as liability shifts towards OEMs, tier 1 and 2 suppliers and software developers. Vehicle accident risks will hinge on vehicular characteristics in addition to driver related risks as drivers alternate between autonomous and manual driving modes. This paper proposes a Bayesian Network statistical risk estimation approach that can accommodate changing risk levels and the emergence of new risk structures. We demonstrate the use of this method for a Level 3 semi-autonomous vehicle for two scenarios, one where the driver is in control and one where the vehicle is in control. This approach is especially suited to use telematics data generated from the vehicle inherent technologies. We validate the efficacy of this approach from the perspective of the insurer and discuss how vehicle technology development will require a greater degree of collaboration between the insurance company and the manufacturers in order to develop a greater understanding of the risks semi-autonomous and fully autonomous vehicles.     

A hybrid Bayesian Network approach to detect driver cognitive distraction

Driver cognitive distraction (e.g., hand-free cell phone conversation) can lead to unapparent, but detrimental, impairment to driving safety. Detecting cognitive distraction represents an important function for driver distraction mitigation systems. We developed a layered algorithm that integrated two data mining methods—Dynamic Bayesian Network (DBN) and supervised clustering—to detect cognitive distraction using eye movement and driving performance measures. In this study, the algorithm was trained and tested with the data collected in a simulator-based study, where drivers drove either with or without an auditory secondary task. We calculated 19 distraction indicators and defined cognitive distraction using the experimental condition (i.e., “distraction” as in the drives with the secondary task, and “no distraction” as in the drives without the secondary task). We compared the layered algorithm with previously developed DBN and Support Vector Machine (SVM) algorithms. The results showed that the layered algorithm achieved comparable prediction performance as the two alternatives. Nonetheless, the layered algorithm shortened training and prediction time compared to the original DBN because supervised clustering improved computational efficiency by reducing the number of inputs for DBNs. Moreover, the supervised clustering of the layered algorithm revealed rich information on the relationship between driver cognitive state and performance. This study demonstrates that the layered algorithm can capitalize on the best attributes of component data mining methods and can identify human cognitive state efficiently. The study also shows the value in considering the supervised clustering method as an approach to feature reduction in data mining applications.     

A dynamic Bayesian network model for real-time crash prediction using traffic speed conditions data

Traffic crashes occurring on freeways/expressways are considered to relate closely to previous traffic conditions, which are time-varying. Meanwhile, most studies use volume/occupancy/speed parameters to predict the likelihood of crashes, which are invalid for roads where the traffic conditions are estimated using speed data extracted from sampled floating cars or smart phones. Therefore, a dynamic Bayesian network (DBN) model of time sequence traffic data has been proposed to investigate the relationship between crash occurrence and dynamic speed condition data. Moreover, the traffic conditions near the crash site were identified as several state combinations according to the level of congestion and included in the DBN model. Based on 551 crashes and corresponding speed information collected on expressways in Shanghai, China, DBN models were built with time series speed condition data and different state combinations. A comparative analysis of the DBN model using flow detector data and a static Bayesian network model was also conducted. The results show that, with only speed condition data and nine traffic state combinations, the DBN model can achieve a crash prediction accuracy of 76.4% with a false alarm rate of 23.7%. In addition, the results of transferability testing imply that the DBN models are applicable to other similar expressways with 67.0% crash prediction accuracy.     

Who you are is how you travel: A framework for transportation mode detection using individual and environmental characteristics

With the increasing prevalence of geo-enabled mobile phone applications, researchers can collect mobility data at a relatively high spatial and temporal resolution. Such data, however, lack semantic information such as the interaction of individuals with the transportation modes available. On the other hand, traditional mobility surveys provide detailed snapshots of the relation between socio-demographic characteristics and choice of transportation modes. Transportation mode detection is currently approached using features such as speed, acceleration and direction either on their own or in combination with GIS data. Combining such information with socio-demographic characteristics of travellers has the potential of offering a richer modelling framework that could facilitate better transportation mode detection using variables such as age and disability. In this paper, we explore the possibility to include both elements of the environment and individual characteristics of travellers in the task of transportation mode detection. Using dynamic Bayesian Networks, we model the transition matrix to account for such auxiliary data by using an informative Dirichlet prior constructed using data from traditional mobility surveys. Results have shown that it is possible to achieve comparable accuracy with the most widely used classification algorithms while having a rich modelling framework, even in the case of sparse mobility data.