Comparing practice-ready forecast models for weekly and monthly fluctuations of average daily traffic and enhancing accuracy by weighting methods

Knowing daily traffic for the current year is recognized as being essential in many fields of transport analysis and practice, and short-term forecasting models offer a set of tools to meet these needs. This paper examines and compares the accuracy of three representative parametric and non-parametric prediction models, selected by the analysis of the numerous methods proposed in the literature for their good combi- nation of forecast accuracy and ease of calibration, using real-life data on Italian motorway stretches. Non-parametric K-NN regression model, Gaussian maximum likelihood model and double seasonality Holt-Winters exponential smoothing model confirm their goodness to predict the weekly and monthly fluctuations of average daily traffic with varying degrees of performance, while maintaining an easy use in professional practice, i.e. requiring ordinary professional skills and conventional analysis tools. Since combining several prediction models can give, on average, more accuracy than that of the individual models, the paper compares two weighting methods of easy implementation and susceptible to a direct use, namely the widely used information entropy method and the less widespread Shapley value method. Despite being less common than the information entropy method, the Shapley value method proves to be more capable in better combining single forecasts and produces improvements in the predictions for test data. With these remarks, the paper might be of interest to traffic technicians or analysts, in various and not uncommon tasks they might find in their work.

A multi-class time-dependent model for the analysis of waiting phenomena at a motorway tollgate

The planning, design and operational management of motorway toll booths are of great interest in traffic engineering, as these facilities directly influence the quality of the service offered to users. This paper focuses on a time-dependent queue model based on the coordinates transformation criterion for operations assessment at a motorway tollgate. This model allows to face the whole spectrum of situations that may characterize a toll booth,some of which often fall outside the boundaries of the probabilistic theory for stationary queues.The paper proposes an M=G=1 multi-class queue model for the evaluation of evolutionary profiles of waiting times and queue lengths by closed-form equations. The results obtained for three numerical test cases show a good approximation level, compared with the mean values of queue parameters obtained reiterating a discrete-state simulation model.The proposed time-dependent equations will be useful in technical cases, allowing to operate quickly and compactly even when probabilistic queue theory is not applicable or produce unrealistic results, and the burden of complexity of the simulation approach is not conveniently absorbable. The discussion highlights a significant flexibility of the model proposed in addressing situations with conventional vehicles, i.e., with total human control and proposes some considerations for application in future scenarios with the presence of connected vehicles(CVs).

State of the art and computational aspects of time-dependent waiting models for non-signalised intersections

Time-dependent models are of great importance in highway engineering as they are appropriate for evaluating waiting times and queue lengths at intersections,which are integral parts of various activities in planning,verification and decision support for infrastructure.After reviewing the literature of the main time-dependent models based on the coordinate transformation method and a discussion about some computational issues in time-evolution profiles for non-signalised intersections,the paper identifies the requirements these models have to satisfy in order to be used as"basic"cases for analysing complex evolutionary situations.Three"basic"cases are presented with their timedependent equations for vehicle waiting times and vehicle number;they have been completed and dimensionally homogenised in this paper.As they are recursive,these formulas can be applied for sequential intervals in the time domain in both vehicles and passenger car units.The closed-form expressions for state variables show to be mutually equivalent in comparison with discrete event simulation models and imbedded Markov chain results.For all the three models the paper presents a common deterministic simplification for average waiting time,with good approximation results in the tested cases.The proposed time-dependent formulas will contribute to a better adherence to the real phenomena,compared to the extremely simplified and unrealistic methods suggested by the international manuals for level of service assessment.The proposed formulas will be useful for current applications and possible future development in order to meet the emerging needs of road and transport engineering.