Search

Predicting transportation mode choice is a classic challenge of travel behavior research. Over the years, different theoretical concepts and modeling approaches have been applied. This paper elaborates the application of class association rules (CARs) and examines their predictive performance using data extracted from the 2015 National Dutch Travel Survey. To solve the problem how to activate rules that have high confidence but low support, the information gain (IG) concept is introduced in the model building process. The modeling process in this study first involves extracting frequent items from the data using the FP-Growth algorithm and deriving CARs from these frequent items. Next, the IG statistic is used to construct a novel model (named CARIG), which consists of a set of decision rules that formally represent behavioral scripts, for predicting individuals’ transportation mode choice. The performance of CARIG is compared with the performance of conventional class-based association rules (CBA), decision trees (DT), a convolutional neural network (CNN) and a logistic regression (LR) model. In addition, a 10-fold cross validation test using a grid search parameter optimization method is conducted to validate the proposed approach. The results show that the proposed method is promising in predicting transportation mode choices observed in the national travel survey data.

Predicting transportation mode choice is a critical component of forecasting travel demand. Recently, machine learning methods have become increasingly more popular in predicting transportation mode choice. Class association rules (CARs) have been applied to transportation mode choice, but the application of the imputed rules for prediction remains a long-standing challenge. Based on CARs, this paper proposes a new rule merging approach, called CARM, to improve predictive accuracy. In the suggested approach, first, CARs are imputed from the frequent pattern tree (FP-tree) based on the frequent pattern growth (FP-growth) algorithm. Next, the rules are pruned based on the concept of pessimistic error rate. Finally, the rules are merged to form new rules without increasing predictive error. Using the 2015 Dutch National Travel Survey, the performance of suggested model is compared with the performance of CARIG that uses the information gain statistic to generate new rules, class-based association rules (CBA), decision trees (DT) and the multinomial logit (MNL) model. In addition, the proposed model is assessed using a ten-fold cross validation test. The results show that the accuracy of the proposed model is 91.1%, which outperforms CARIG, CBA, DT and the MNL model.

To develop effective strategies for the supply of shared parking and study various theoretical choice models under uncertainty, this paper investigates private parking space owners’ propensity to engage in shared parking schemes using a stated choice experiment that involves an uncertain key attribute. A hybrid expected utility-regret model incorporating rejoice is specified to explore the participation behavior. Equivalent models considering the perception of attribute differences are also estimated. Results show that socio-demographic characteristics, social influence, government’s role, media attention, platform fee, and revenues are all important factors explaining private parking owners’ propensity to engage in shared parking schemes. Besides, the model incorporating all these components, including the emotions of regret and rejoice and the perception of attribute differences, yields the best results. These findings could help promote the policy development toward increasing people’s engagement in shared parking.

Mobility-as-a-Service (MaaS) can be considered the latest innovative technological solution to sustainable transportation. It offers travelers access to a pre-paid bundle of transportation modes on a single app with the additional convenience of the integration of planning, booking, and payment. Capturing the impact of this new technology on travel behavior remains a challenge, given the scarce number of comprehensive MaaS pilots. In this study, a stated adaptation experiment was designed to determine travelers’ propensity toward using this new service. Such knowledge sheds light on the potential of MaaS to alter daily travel patterns.

Route choice modeling under travel time uncertainty is essential for analyzing travelers’ mobility patterns. This paper investigates the impact of travel time uncertainty on route choice behavior in user equilibrium models based on a generalized mean-variance metric (GMV). This model can capture the influence of risk attitudes and schedule unpunctuality on route choice using a generalization of expected travel time, variance, and expected early or late arrival penalties, of which travelers are assumed to minimize the GMV of trips considering a certain on-time arrival probability. This paper establishes the properties of GMV and formulates the GMV-based static user equilibria as a variational inequality (VI) problem, for which the existence and uniqueness of the solutions are also analyzed. An effective traffic assignment algorithm without path enumeration is developed to solve the proposed user equilibrium problem. Numerical examples are conducted to demonstrate the properties of the proposed model. KEYWORDS: Travel time uncertainty, route choice, generalize mean-variance, user equilibrium

This paper presents an integrated framework for the optimal planning of public charging stations for plug-in electric vehicles (PEVs) in urban areas. The framework consists of two main components: (i) an out-of-home charging demand model based on an activity-based travel demand model, and (ii) a public charging station location-allocation model using a scenario-based stochastic programming (SP) approach. In order to capture the dynamic charging behaviour of PEV users, a chi-squared automatic interaction detector (CHAID)-based mixed effects decision tree is induced from multi-day activity diaries. Moreover, because the stochastic error of the micro-simulation approach brings about uncertainty, we adopted a two-stage stochastic mixed-integer programming (TSMIP) model, which measures uncertainty by means of a finite set of scenarios obtained from the derived decision rules underlying PEV charging. The proposed approach is demonstrated for the City of Eindhoven, The Netherlands, and benefits of the stochastic solution are discussed.

There is ample evidence that individuals’ evaluation and choice of transportation-related alternatives in stated choice experiments vary by task complexity. Previous research has analyzed experimental error primarily as a function of the number of choice alternatives, the number of attribute (levels) and response time. It has ignored the fact that the cognitive ability of individuals may be another important moderating variable. Another limitation of previous research is that experimental error has been studied exclusively for utility-maximizing models, not for the more recently introduced regret-rejoice models. To augment this body of research, this study therefore proposes two advanced flexible reference-dependence models, based on the concepts of regret and rejoice, and argues that information processing depends on task complexity, which is affected not only by information load and response time but also by subjects’ cognitive ability. We apply an integrated approach examining both varying the variance of error term and simplifying task complexity by considering only partial information. Furthermore, we explore how the effect of information load differs between long-term and short-term decision problems. To that end, two stated choice experiments with varying information load are designed, one related to a short-term decision context (route choice) and one about a long-term decision context (residential choice). Cognitive ability is explicitly measured using a validated psychological test. Results show subjects’ inclination to consider full information in arriving at a choice increases with increasing cognitive ability. Response time and age appeared to be other significant determinants for the magnitude of error variance of choices. The proposed models outperform the conventional utility model and conventional regret- rejoice models and result in less error in the prediction of market share. The improvement in the prediction performance is more pronounced for mo

Multi-state supernetworks are capable of representing activity-travel patterns at a high level of detail and thus are a powerful tool for activity-travel scheduling (ATS) of multidimensional choice facets. To alleviate the limitations of the common deterministic network representation, travel time and activity duration uncertainty has been incorporated in multi-state supernetworks. However, the extension unrealistically assumed that all uncertain components are independent. This study suggests an approach of ATS considering spatially and temporally correlated travel times and activity durations in stochastic time-dependent (STD) multi-state supernetworks. Support points are used as a representation of the stochasticity of the STD multi-state supernetwork. ATS is formulated as a pathfinding problem subject to space–time constraints based on recursive formulations. A series of numerical experiments is implemented to demonstrate the applicability of the suggested approach.

This paper follows authors’ previous work, which measured social influence in travel behavior using a sequential stated adaptation experiment, and aims to investigate issues that are not discussed in the previous one. Specifically, this paper supplements the previous work in two aspects. On the one hand, the previous work tried to model respondents’ sequential choice in terms of a choice task with and without social network member’s choice, and estimated the model sequentially without any clear evidences to show the difference between sequential and simultaneous estimation. Therefore, this paper first provides the comparison of these two estimation approaches. On the other hand, the previous work did not clearly address two specific issues that come with the use of sequential choice experiments to measure social influence. The first issue is that respondents have a tendency to keep their choices consistent in sequential choice tasks, which may mask the effect of social influence; the second one is that the stochastic nature of utility implies that respondents may still reveal a shift in their choices even if no information about the choice of social network members is provided, which will be mixed with the effect of social influence, depending on the similarity of alternatives in the choice set. Therefore, this paper aims to take these two issues into consideration. The results of this study reveal that estimating the model simultaneously may lead to confounding bias and that taking the issues caused by the sequential choice experiment could improve the performance of the model and provide more insights about social influence.