Local online kernel ridge regression for forecasting of urban travel times

Accurate and reliable forecasting of traffic variables is one of the primary functions of Intelligent Transportation Systems. Reliable systems that are able to forecast traffic conditions accurately, multiple time steps into the future, are required for advanced traveller information systems. However, traffic forecasting is a difficult task because of the nonlinear and nonstationary properties of traffic series. Traditional linear models are incapable of modelling such properties, and typically perform poorly, particularly when conditions differ from the norm. Machine learning approaches such as artificial neural networks, nonparametric regression and kernel methods (KMs) have often been shown to outperform linear models in the literature. A bottleneck of the latter approach is that the information pertaining to all previous traffic states must be contained within the kernel, but the computational complexity of KMs usually scales cubically with the number of data points in the kernel. In this paper, a novel kernel-based machine learning (ML) algorithm is developed, namely the local online kernel ridge regression (LOKRR) model. Exploiting the observation that traffic data exhibits strong cyclic patterns characterised by rush hour traffic, LOKRR makes use of local kernels with varying parameters that are defined around each time point. This approach has 3 advantages over the standard single kernel approach: (1) It allows parameters to vary by time of day, capturing the time varying distribution of traffic data; (2) It allows smaller kernels to be defined that contain only the relevant traffic patterns, and; (3) It is online, allowing new traffic data to be incorporated as it arrives. The model is applied to the forecasting of travel times on London’s road network, and is found to outperform three benchmark models in forecasting up to 1 h ahead.     

A mixed Bayesian network for two-dimensional decision modeling of departure time and mode choice

The modeling of travel decision making has been a popular topic in transportation planning. Previous studies focused on random-utility discrete choice models and machine learning methods. This paper proposes a new modeling approach that utilizes a mixed Bayesian network (BN) for travel decision inference. The authors use a predetermined BN structure and calculate priori and posterior probability distributions of the decision alternatives based on the observed explanatory variables. As a “utility-free” decision inference method, the BN model releases the linear structure in the utility function but assumes the traffic level of service variables follow multivariate Gaussian distribution conditional on the choice variable. A real-world case study is conducted by using the regional travel survey data for a two-dimensional decision modeling of both departure time choice and travel mode choice. The results indicate that a two-dimensional mixed BN provides better accuracy than decision tree models and nested logit models. In addition, one can derive continuous elasticity with respect to each continuous explanatory variable for sensitivity analysis. This new approach addresses a research gap in probabilistic travel decision making modeling as well as two-dimensional travel decision modeling.     

Modelling traffic flows and estimating road travel times in transportation network under dynamic disturbances

Transportation - Traffic congestion is a common phenomenon in road transportation networks, especially during peak hours. More accurate prediction of dynamic traffic flows is very important for...     

Synthesis of first practices and operational research approaches in activity-based travel demand modeling

Regional travel models in the United States are clearly evolving from conventional models towards a new generation of more behaviorally realistic activity-based models. The new generation of regional travel demand models is characterized by three features: (1) an activity-based platform, that implies that modeled travel be derived within a general framework of the daily activities undertaken by households and persons, (2) a tour-based structure of travel where the tour is used as the basic unit of modeling travel instead of the elemental trip, and (3) micro-simulation modeling techniques that are applied at the fully-disaggregate level of persons and households, which convert activity and travel related choices from fractional-probability model outcomes into a series of discrete or “crisp” decisions.While the new generation of model has obvious conceptual advantages over the conventional four-step models, there are still numerous technical issues that have to be addressed as well as a better understanding of practical benefits should be achieved before the new generation of models can fully replace conventional models. The paper summarizes the recent successful experience in the development and application of activity-based demand models for Metropolitan Planning Organizations in the US. Moving activity-based approaches into practice is analyzed in a broad context of travel demand modeling market tendencies and policy implications.     

Bayesian committee of neural networks to predict travel times with confidence intervals

Short-term prediction of travel time is one of the central topics in current transportation research and practice. Among the more successful travel time prediction approaches are neural networks and combined prediction models (a ‘committee’). However, both approaches have disadvantages. Usually many candidate neural networks are trained and the best performing one is selected. However, it is difficult and arbitrary to select the optimal network. In committee approaches a principled and mathematically sound framework to combine travel time predictions is lacking. This paper overcomes the drawbacks of both approaches by combining neural networks in a committee using Bayesian inference theory. An ‘evidence’ factor can be calculated for each model, which can be used as a stopping criterion during training, and as a tool to select and combine different neural networks. Along with higher prediction accuracy, this approach allows for accurate estimation of confidence intervals for the predictions. When comparing the committee predictions to single neural network predictions on the A12 motorway in the Netherlands it is concluded that the approach indeed leads to improved travel time prediction accuracy.     

A method to assess demand growth vulnerability of travel times on road network links

Many national governments around the world have turned their recent focus to monitoring the actual reliability of their road networks. In parallel there have been major research efforts aimed at developing modelling approaches for predicting the potential vulnerability of such networks, and in forecasting the future impact of any mitigating actions. In practice—whether monitoring the past or planning for the future—a confounding factor may arise, namely the potential for systematic growth in demand over a period of years. As this growth occurs the networks will operate in a regime closer to capacity, in which they are more sensitive to any variation in flow or capacity. Such growth will be partially an explanation for trends observed in historic data, and it will have an impact in forecasting too, where we can interpret this as implying that the networks are vulnerable to demand growth. This fact is not reflected in current vulnerability methods which focus almost exclusively on vulnerability to loss in capacity. In the paper, a simple, moment-based method is developed to separate out this effect of demand growth on the distribution of travel times on a network link, the aim being to develop a simple, tractable, analytic method for medium-term planning applications. Thus the impact of demand growth on the mean, variance and skewness in travel times may be isolated. For given critical changes in these summary measures, we are thus able to identify what (location-specific) level of demand growth would cause these critical values to be exceeded, and this level is referred to as Demand Growth Reliability Vulnerability (DGRV). Computing the DGRV index for each link of a network also allows the planner to identify the most vulnerable locations, in terms of their ability to accommodate growth in demand. Numerical examples are used to illustrate the principles and computation of the DGRV measure.     

Decision tree method for modeling travel mode switching in a dynamic behavioral process

As road congestion is exacerbated in most metropolitan areas, many transportation policies and planning strategies try to nudge travelers to switch to other more sustainable modes of transportation. In order to better analyze these strategies, there is a need to accurately model travelers’ mode-switching behavior. In this paper, a popular artificial intelligence approach, the decision tree (DT), is used to explore the underlying rules of travelers’ switching decisions between two modes under a proposed framework of dynamic mode searching and switching. An effective and practical method for a mode-switching DT induction is proposed. A loss matrix is introduced to handle class imbalance issues. Important factors and their relative importance are analyzed through information gains and feature selections. Household Travel Survey data are used to implement and validate the proposed DT induction method. Through comparison with logit models, the improved prediction ability of the DT models is demonstrated.     

Psychological items: a useful addition in modeling travel behavior on managed lanes

Transportation - Managed lanes (MLs) are a tool to more efficiently operate segments of a freeway. As ML prevalence increases in the United States of America, it is important to understand travel...     

Specification,transferability and the effect of data outliers in modeling the choice of mode in urban travel

Three problems of great importance to urban travel demand modeling using multinominal logit models are examined in this paper. They are (1) the effect of data outliers on model coefficients; (2) the effect of model specification on coefficients and model explanatory power; and (3) the transferability of model coefficients within the region, between regions, and in time.Four data sets are used in the study. They are: Washington, D.C., Minneapolis-St. Paul, and two data sets from the San Francisco Bay Area, Pre-BART and Post-BART. The data are standard home-interview survey data appended with network supplied modal travel cost and time information.The findings of the research are occasionally contradictory; the majority of the evidence supports the following conclusions. The outliers do not have a statistically significant effect (at 0.05 level) on the coefficients; however, the outliers can have a substantial effect on the point estimates of some of the coefficients. Model specification has an impact on model coefficients and model explanatory power. In particular, the definition of out-of-vehicle travel time appears to be important and, if available, the use of separate walk and wait times is preferred over their sum, the out-of-vehicle time. Finally, the model coefficients do not appear transferable within region, between regions, or in time.Research was supported in part by the Alfred P. Sloan Foundation, through grant 74-12-8 to the Department of Economics, University of California, Berkeley and by the National Science Foundation, through grant APR 74-20392, Research Applied to National Needs Program, to the University of California, Berkeley.     

Misclassification in travel surveys and implications to choice modeling: application to household auto ownership decisions

Travel surveys that elicit responses to questions regarding daily activity and travel choices form the basis for most of the transportation planning and policy analysis. The response variables collected in these surveys are prone to errors leading to mismeasurement or misclassification. Standard modeling methods that ignore these errors while modeling travel choices can lead to biased parameter estimates. In this study, methods available in the econometrics literature were used to quantify and assess the impact of misclassification errors in auto ownership choice data. The results uncovered significant misclassification rates ranging from 1 to 40% for different auto ownership alternatives. Also, the results from latent class models provide evidence for variation in misclassification probabilities across different population segments. Models that ignore misclassification were not only found to have lower statistical fit but also significantly different elasticity effects for choice alternatives with high misclassification probabilities. The methods developed in this study can be extended to analyze misclassification in several response variables (e.g., mode choice, activity purpose, trip/tour frequency, and mileage) that constitute the core of advanced travel demand models including tour and activity-based models.

     

Impacts of highly automated vehicles on travel demand: macroscopic modeling methods and some results

Automated vehicles (AV) will change transport supply and influence travel demand. To evaluate those changes, existing travel demand models need to be extended. This paper presents ways of integrating characteristics of AV into traditional macroscopic travel demand models based on the four-step algorithm. It discusses two model extensions. The first extension allows incorporating impacts of AV on traffic flow performance by assigning specific passenger car unit factors that depend on roadway type and the capabilities of the vehicles. The second extension enables travel demand models to calculate demand changes caused by a different perception of travel time as the active driving time is reduced. The presented methods are applied to a use case of a regional macroscopic travel demand model. The basic assumption is that AV are considered highly but not fully automated and still require a driver for parts of the trip. Model results indicate that first-generation AV, probably being rather cautious, may decrease traffic performance. Further developed AV will improve performance on some parts of the network. Together with a reduction in active driving time, cars will become even more attractive, resulting in a modal shift towards car. Both circumstances lead to an increase in time spent and distance traveled.

     

Behavioral modeling of on-demand mobility services: general framework and application to sustainable travel incentives

Transportation - This paper presents a systematic way of understanding and modeling traveler behavior in response to on-demand mobility services. We explicitly consider the sequential and yet...     

Traffic congestion and economic context: changes of spatiotemporal patterns of traffic travel times during crisis and post-crisis periods

Transportation - This paper aims to evaluate the impacts of the economic context on traffic congestion and its consequential effects on private vehicle accessibility. We conduct a long-term...     

The measure of all things: reflections on changing conceptions of the individual in travel demand modeling

Transportation demand modeling has evolved in scope, theory, and practice in the many decades since the US Bureau of Public Roads pioneered home interview transportation studies in metropolitan households in the early 1940s. The major currents of these developments are discussed in the present paper through consideration of the changing role of the individual—as a source of data, as a unit of analysis, and as the intended beneficiary. In addition to reviewing this history we raise, but are unable to resolve, a growing current concern, namely how the public interest can be best served when the transportation data of greatest value is collected by private entities.     

Inter-temporal variation in the travel time and travel cost parameters of transport models

The parameters for travel time and travel cost are central in travel demand forecasting models. Since valuation of infrastructure investments requires prediction of travel demand for future evaluation years, inter-temporal variation of the travel time and travel cost parameters is a key issue in forecasting. Using two identical stated choice experiments conducted among Swedish drivers with an interval of 13 years, 1994 and 2007, this paper estimates the inter-temporal variation in travel time and cost parameters (under the assumption that the variance of the error components of the indirect utility function is equal across the two datasets). It is found that the travel time parameter has remained constant over time but that the travel cost parameter has declined in real terms. The trend decline in the cost parameter can be entirely explained by higher average income level in the 2007 sample compared to the 1994 sample. The results support the recommendation to keep the travel time parameter constant over time in forecast models, but to deflate the travel cost parameter with the forecasted income increase among travellers and the relevant income elasticity of the cost parameter. Evidence from this study further suggests that the inter-temporal and the cross-sectional income elasticities of the cost parameter are equal. The average elasticity is found to be ?0.8 to ?0.9 in the present sample of drivers, and the elasticity is found to increase with the real income level, both in the cross-section and over time.     

Evaluation of a cellular phone-based system for measurements of traffic speeds and travel times: A case study from Israel

The purpose of this paper is to examine the performance of a new operational system for measuring traffic speeds and travel times which is based on information from a cellular phone service provider. Cellular measurements are compared with those obtained by dual magnetic loop detectors. The comparison uses data for a busy 14 km freeway with 10 interchanges, in both directions, during January–March of 2005. The dataset contains 1 284 587 valid loop detector speed measurements and 440 331 valid measurements from the cellular system, each measurement referring to a 5 min interval. During one week in this period, 25 floating car measurements were conducted as additional comparison observations. The analyses include visual, graphical, and statistical techniques; focusing in particular on comparisons of speed patterns in the time–space domain. The main finding is that there is a good match between the two measurement methods, indicating that the cellular phone-based system can be useful for various practical applications such as advanced traveler information systems and evaluating system performance for modeling and planning.     

Heterogeneity in departure time preferences,flexibility and schedule constraints

Transportation - This study develops a latent class choice model of departure time preferences for morning commute trips by car. The model is empirically evaluated using a sample of car commuters...     

Using structural equations modeling to unravel the influence of land use patterns on travel behavior of workers in Montreal

This paper addresses the relations between travel behavior and land use patterns using a Structural Equations Modeling (SEM) framework. The proposed model structure draws on two earlier models developed for Lisbon and Seattle which show significant effects of land use patterns on travel behavior. The travel behavior variables included here are multifaceted including commuting distance, car ownership, the amount of mobility by mode (car, transit and non-motorized modes), both in terms of total kilometers travelled and number of trips. The model also includes a travel scheduling variable, which is the total time spent between the first and last trips to reflect daily constraints in time allocation and travel.The modeled land use variables measure the levels of urban concentration and density, diversity, both in terms of types of uses and the mix between jobs and inhabitants/residents, the transport supply levels, transit and road infrastructure, and accessibility indicators. The land use patterns are described both at the residence and employment zones of each individual included in the model by using a factor analysis technique as a data reduction and multicollinearity elimination technique. In order to explicitly account for self selection bias the land use variables are explicitly modeled as functions of socioeconomic attributes of individuals and their households.The results obtained show that people with different socioeconomic characteristics tend to work and live in places of substantially different urban environments. But besides these socioeconomic self-selection effects, land use variables significantly affect travel behavior. More precisely the effects of land use are in great part passed thru variables describing long term decisions like commuting distance, and car ownership. These results point to similar conclusions from the models developed for Lisbon and Seattle and thus give weight to the use of land use policies as tools for changing travel behavior.     

Error and uncertainty in travel surveys

Transportation study home interview surveys normally underestimate the amount of trip-making by residents of a study area, and the data has to be adjusted subsequently to compensate for this effect. One important reason for the shortfall appears to be the incomplete recall and reporting of travel information by respondents. The paper examines the influence of the survey instrument on recall by comparing reported travel behaviour in two surveys conducted in the same town, one using a conventional travel diary and the other an activity diary. The latter format appears to encourage a more complete response, giving rise to significantly higher reported trip rates and travel times. The paper discusses some research and policy implications of this finding and concludes with tentative suggestions for improved travel survey instruments.