Zeitschrift fur Wahrscheinlichkeitstheorie und Verwandte Gebiete.

We present BusTr, a machine-learned model for translating road traffic forecasts into predictions of bus delays, used by Google Maps to serve the majority of the world's public transit systems where no official real-time bus tracking is provided. We demonstrate that our neural sequence model improves over DeepTTE, the state-of-the-art baseline, both in performance (-30% MAPE) and training stability. We also demonstrate significant generalization gains over simpler models, evaluated on longitudinal data to cope with a constantly evolving world.

BusTr: Predicting Bus Travel Times from Real-Time Traffic

https://dl.acm.org/doi/pdf/10.1145/3394486.3403376

This chapter presents the Internet of Vehicles (IoV) concept, technologies and applications used to realise intelligent functions, optimise vehicle performance, control, and decision-making for future electric, connected, autonomous, and shared (ECAS) vehicles mobility scenarios. The concept addresses the convergence of the edge intelligence embedded in the vehicles based on Artificial Intelligence (AI) technologies with the cooperative, collaborative intelligence distributed into the Internet of Things (IoT) devices and edge computing infrastructure federated with the hierarchical cognitive processes and analytics in physical, network, infrastructure, and data spaces. IoV integrates and links the ECAS vehicles’ domains with the intra-vehicle networks, vehicle-to-everything (V2X) networks, the processing and cognitive functions provided by federated platforms using the intelligent edge and cloud computing infrastructure. The cognitive transformation of vehicles and the integration of immersive technologies, i.e., virtual reality (VR) and augmented reality (AR) combined with intelligent connectivity, IoT, Distributed Ledger Technologies (DLTs), digital twins (DTs), data, knowledge, security/privacy requirements and learning platforms provide the technological foundation of IoV allowing for entirely new services, applications, and user experiences. The paper advances the latest architectural concepts for ECAS vehicles. It proposes an IoV 3D multi-layered architecture that combines AI, edge computing and connectivity as part of the functional layers while integrating the system properties and trustworthiness properties into the overall architecture to provide efficient new mobility applications and services. The proposed system of systems concept for IoV applications allows for distributed intelligent functions to be embedded into the edge and cloud infrastructure for ECAS vehicles to provide a computing, processing, and intelligent connectivity continuum for IoV applications and services.

Internet of Vehicles – System of Systems Distributed Intelligence for Mobility Applications

The arrival of a ship at a port is a complex process. It has been reported that ships in port can spend 5–10% of their time in unnecessary anchorage. Predicting the accurate time of ship arrivals could support port actors to optimally allocate their resources and reduce this time, to the benefit of port competitiveness. Currently, the rise of real-time automated ship identification data has given ports, shipping lines and other actors the possibility to predict but also to follow the consecutive steps a ship undertakes while entering a port. Much work has been done on the development of advanced estimated time of arrival (ETA) predictors, but little is known about the quality of ETA information that shipping lines provide to ports. We explore, quantitatively, the predictive power of the sequences of ETAs that ships transmit to the Port of Antwerp while arriving. We propose several forecast evaluation measures to assess the quality of a sequence of predictions and their ability to converge to the actual time of arrival. For the Port of Antwerp, we compute and interpret the measures on a dataset of port calls in the year 2019. Our aim is to compare ETA prediction quality across ships. Understanding variations in prediction quality gives port authorities a basis for discussion with individual shipowners, enabling them to improve their predictions and thus contribute to the competitive performance of the port. As an innovation, we cluster the different berths in our data set with the OPTICS algorithm to gain a better understanding of the effects of the river locks a ship must transit on her way to the allocated berth. Our findings indicate that, on average, shipping lines offer poor and optimistic arrival estimates, which they gradually adjust. The adjustments increase as the ship approaches the port and sometimes they overcompensate earlier predictions. In this, we find little evidence that shipping lines are adopting the advanced arrival prediction methods that have been developed in recent years. Our proposed prediction evaluation measures open the door for a uniform and quantitative approach to assessing the quality of prediction sequences of ships entering ports, to the benefit of a miscellany of port actors.

On the quality of ship arrival predictions

The majority of public transport vehicles are fitted with Automatic Vehicle Location (AVL) systems generating a continuous stream of data. The availability of this data has led to a substantial body of literature addressing the development of algorithms to predict Estimated Times of Arrival (ETA). Here research literature reporting the development of ETA prediction systems specific to busses is reviewed to give an overview of the state of the art. Generally, reviews in this area categorise publications according to the type of algorithm used, which does not allow an objective comparison. Therefore this survey will categorise the reviewed publications according to the input data used to develop the algorithm. The review highlighted inconsistencies in reporting standards of the literature. The inconsistencies were found in the varying measurements of accuracy preventing any comparison and the frequent omission of a benchmark algorithm. Furthermore, some publications were lacking in overall quality. Due to these highlighted issues, any objective comparison of prediction accuracies is impossible. The bus ETA research field therefore requires a universal set of standards to ensure the quality of reported algorithms. This could be achieved by using benchmark datasets or algorithms and ensuring the publication of any code developed.

Survey of ETA prediction methods in public transport networks

This study presents a working concept of a model architecture allowing to leverage the state of an entire transport network to make estimated arrival time (ETA) and next-step location predictions. To this end, a combination of an attention mechanism with a dynamically changing recurrent neural network (RNN)-based encoder library is used. To achieve this, an attention mechanism was employed that incorporates the states of other vehicles in the network by encoding their positions using gated recurrent units (GRUs) of the individual bus line to encode their current state. By muting specific parts of the imputed information, their impact on prediction accuracy can be estimated on a subset of the available data. The results of the experimental investigation show that the full model with access to all the network data performed better in some scenarios. However, a model limited to vehicles of the same line ahead of the target was the best performing model, suggesting that the incorporation of additional data can have a negative impact on the prediction accuracy if they do not add any useful information. This could be caused by poor data quality but also by a lack of interaction between the included lines and the target line. The technical aspects of this study are challenging and resulted in a very inefficient training procedure. We highlight several areas where improvements to our presented method are required to make it a viable alternative to current methods. The findings in this study should be considered as a possible and promising avenue for further research into this novel architecture. As such, it is a stepping stone for future research to improve public transport predictions if network operators provide high-quality datasets.

/pdf/a-model-architecture-for-public-transport-networks-using-a-2fi8sr9g.pdf

A Model Architecture for Public Transport Networks Using a Combination of a Recurrent Neural Network Encoder Library and a Attention Mechanism

The introduction of electronic patient records in the ambulance service provides new opportunities to monitor the population. Most patients presenting to British ambulance services are discharged at scene. Ambulance records are therefore an ideal data source for syndromic early event detection systems to monitor infectious disease in the prehospital population. It has been previously found that tympanic temperature records can be used to detect influenza outbreaks in emergency departments. This study investigated whether routine tympanic temperature readings collected by ambulance crews can be used to detect seasonal influenza. Here we show that these temperature readings do allow the detection of seasonal influenza before methods applied to conventional data sources. The counts of pyretic patients were used to calculate a sliding case ratio (CR) as a measurement to detect seasonal influenza outbreaks. This method does not rely on conventional thresholds and can be adapted to the data. The data collected correlated with seasonal influenza. The 2016/17 outbreak was detected with high specificity and sensitivity, up to 9 weeks before other surveillance programs. An unanticipated outbreak of E. coli was detected in the same dataset. Our results show that ambulance records can be a useful data source for biosurveillance systems. Two outbreaks caused by different infectious agents have been successfully detected. The routine ambulance records allowed to use tympanic temperature readings that can be used as surveillance tool for febrile diseases. Therefore, this method is a valuable addition to the current surveillance tools.

/pdf/a-proof-of-concept-for-a-syndromic-surveillance-system-based-48emi9bj21.pdf

A Proof Of Concept For A Syndromic Surveillance System Based On Routine Ambulance Records In The South-west Of England, For The Influenza Season 2016/17

Background Digital patient records in the ambulance service have opened up new opportunities for prehospital care. Previously it was demonstrated that prehospital pyrexia numbers are linked to an increase in overall calls to the ambulance service. This study aims to predict the future number of calls using deep-learning methods. Methods Temperature readings for 280,447 patients were generously provided by the South Western Ambulance Service Trust. The data covered the time between 05/01/2016 and 30/04/2017 with overall 44,472 patients being pyretic. A rolling window of 10 days was applied to daily sums for both pyretic and apyretic patients. These windows were used as input features to train machine-learning algorithms predicting the number of calls 10 days ahead. Algorithms tested include Linear Regression (LR), basic Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures. A genetic approach was used to optimise the architecture, in which parameters were randomly modified and over several generations the best performing algorithm will be selected to be further manipulated. To assess performance the Mean Average Percentage Error (MAPE) was used. Results The initial analysis showed that the total patient number and pyretic patient numbers are correlated. The best performing algorithms with varying numbers of hidden units had the following MAPE in comparison to simple LR: LR=19.4%, LSTM (104 units) = 6.1%, RNN (79 units)=6.01%, GRU (80 units)=5.97%. Conclusions These preliminary results suggest that deep-learning methods allow to predict the variations in total number of calls caused by circulating infections. Further investigations will aim to confirm these findings. Once fully verified these algorithms could play a major role in operational planning of any ambulance service by predicting increases in demand.

PP15 Predicting variations of calls to an ambulance service in the UK caused by circulating infections using-deep learning methods

Passengers of urban bus networks often rely on forecasts of Estimated Times of Arrival (ETA) and live-vehicle movements to plan their journeys. ETA predictions are unreliable due to the lack of good quality historical data, while ‘live’ positions in mobile apps suffer from delays in data transmission. This study uses deep neural networks to predict the next position of a bus under various vehicle-location data-quality regimes. Additionally, we assess the effect of the target representation in the prediction problem by encoding it either as unconstrained geographical coordinates, progress along known trajectory or ETA at the next two stops. We demonstrate that without data cleaning, model predictions give false confidence if mean errors are used, highlighting the importance of a holistic assessment of the results. We show that target representation affects the prediction accuracy, by constraining the prediction space. The literature is vague about quality issues in public transport data. Here we show that noisy data is a problem and discuss simple but effective approaches to address these issues. Research generally only focuses on a single method of target representation. Therefore, comparing several methods is a useful addition to the literature. This gives insight into the value of addressing data quality issues in urban transport data to enable better predictions and improve the passenger experience. We show that ‘rephrasing’ the prediction problem by changing the target representation can yield massively improved predictions. Our findings enable researchers using deep learning approaches in public transport to make more informed decisions about essential data cleaning steps and problem representation for improved results.

Thilo Reich

Papers

Survey of ETA prediction methods in public transport networks

A Model Architecture for Public Transport Networks Using a Combination of a Recurrent Neural Network Encoder Library and a Attention Mechanism

A Proof Of Concept For A Syndromic Surveillance System Based On Routine Ambulance Records In The South-west Of England, For The Influenza Season 2016/17

PP15 Predicting variations of calls to an ambulance service in the UK caused by circulating infections using-deep learning methods

Impact of Data Quality and Target Representation on Predictions for Urban Bus Networks