Traditional Multinomial Logit Model Research Articles

Public transport mode choice behavior of different-income passengers during large-scale public health incidents

Public transport ridership has been hit hard by the COVID-19 pandemic in many countries. Investigating passengers' public transport mode choice behavior during large-scale public health incidents can uncover the major influential factors and help propose policies and strategies to reduce the pandemic transmission and recover the public transport revenue. This study develops an integrated choice and latent variables (ICLV) model by income based on structural equation model to model passengers' public transport choice behavior during the normalized stage of the pandemic. The model considers passengers' socioeconomic attributes, travel attributes, and attitude-perception attributes, and can appropriately capture passengers' psychological latent attributes. Taking Beijing China as an example, we collect some revealed preference survey data online. The modeling results show that the risk perception as a mediator variable has a significant impact on mode preference. Moreover, the convenience of public transport has the largest influence on risk perception. These findings suggest that risk perception and the convenience of public transport play a major role in passengers' mode choice behavior. In addition, the impacts of the various influential factors on the public transport mode choices are significantly different across different income groups. Further, the ICLV model can achieve better performance and is superior to the traditional Multinomial Logit model. The modeling framework can help propose targeted and instructive strategies during the normalized stage of the pandemic by uncovering the major influential factors in passengers’ public transport mode choices, which is applicable to similar pandemics in the future.

Predicting travel mode choice with a robust neural network and Shapley additive explanations analysis

AbstractPredicting and understanding travellers’ mode choices is crucial to developing urban transportation systems and formulating traffic demand management strategies. Machine learning (ML) methods have been widely used as promising alternatives to traditional discrete choice models owing to their high prediction accuracy. However, a significant body of ML methods, especially the branch of neural networks, is constrained by overfitting and a lack of model interpretability. This study employs a neural network with feature selection for predicting travel mode choices and Shapley additive explanations (SHAP) analysis for model interpretation. A dataset collected in Chengdu, China was used for experimentation. The results reveal that the neural network achieves commendable prediction performance, with a 12% improvement over the traditional multinomial logit model. Also, feature selection using a combined result from two embedded methods can alleviate the overfitting tendency of the neural network, while establishing a more robust model against redundant or unnecessary variables. Additionally, the SHAP analysis identifies factors such as travel expenditure, age, driving experience, number of cars owned, individual monthly income, and trip purpose as significant features in our dataset. The heterogeneity of mode choice behaviour is significant among demographic groups, including different age, car ownership, and income levels.

Sensitivity evaluation of machine learning-based calibrated transportation mode choice models: A case study of Alexandria City, Egypt

Intelligent methods including Machine Learning (ML) techniques have been increasingly employed in transportation mode choice modeling, which is more complex than other demand models, since it has to reliably and accurately reflect a wide range of related categorical and continuous variables, concerning the travelers, transportation system, and trip characteristics. ML techniques can capture such complex relationships. So that, they can provide a more nuanced understanding of the travelers’ decision process. Most research studies focused mainly on the evaluation of the model accuracy, where little has been done to evaluate the models’ performance toward transport attributes. This research aims to calibrate the transportation choice models using ML techniques, then conduct a comparison with the Multinomial Logit (MNL) model to identify the impact on the model accuracy and performance and quantify to what extent the ML models are sensitive to transport policies compared to the traditional MNL model. To this end, eight ML classifiers were examined. As a case study, the models were calibrated to reflect the choice behavior of trip makers in Alexandria City, Egypt. The models were successfully calibrated with satisfying accuracy; however, the ML models have better calibration results in terms of predictability, outperforming the MNL model, where the GBDT classifier records the best prediction accuracy. Finally, the sensitivity analysis test was performed to quantify the elasticity of the models to transport policies. The results show the ML models’ structure is more comprehensively and accurately built than the MNL model providing better indicative and reliable sensitivity analysis results.

General Deep Multinomial Logit Model

Multinomial logit model (MNL) is by far the most widely used discrete choice model that is widely used to explain or predict a choice from a set of two or more discrete alternatives. MNL operates within a framework of the random utility model (RUM) in which the utility of an alternative perceived by an individual consists of two components: systematic component and random component. The systematic component is usually defined as a linear function. However, practical decision processes involve complex considerations regarding various aspects of the alternatives and individual which cannot be adequately represented by simple linear models. To overcome the weakness of linear utility model and improve the performance of MNL, in this paper, we propose a general deep multinomial logit model (GDMNL) that takes advantage of both traditional MNL and deep learning. In this model, deep neural networks are applied to extend MNL by learning different nonlinear utility functions of various alternatives. The empirical study in the domain of transit route choice analysis demonstrates the validity and superiority of the proposed model.

How does the preference heterogeneity affect the elderly's evaluation of bus accessibility? Evidence from Guangzhou, China

IntroductionThe elderly population in China relies heavily on public transit. The ability of the elderly to access public transit services is linked to their social participation. However, public transit does not address many of the needs of the elderly, and research on the accessibility of public transit for the elderly is insufficient, especially in developing countries such as China. The novelty of this study is that it analyzes the elderly's demographic characteristics, latent variables, and heterogeneity in order to evaluate the accessibility of public transit. It also deepens an understanding of the elderly's travel behavior. MethodsThe primary data for this study were drawn from a survey in Guangzhou, China. A confirmatory factor analysis was used to obtain the elderly's latent variables. After that, a latent-class logit model (LCM) was used to analyze the elderly's travel behavior. LCM avoids the shortcomings of the traditional multinomial logit model (MNL), and avoids shortcomings when segmenting travelers with arbitrary demographic characteristics. ResultsUsing the latent-class logit model, the elderly could be classified into two latent classes: the walking-preference subgroup and the walking-adverse subgroup. These two latent classes accounted for 38.3% and 61.7% of the total interviewed subjects, respectively. Taking in-vehicle time as a reference, it was found that 1 min of walking time was equivalent to 0.36 and 2.82 min of in-vehicle time for these two latent classes, respectively. This reflected a significant difference. The study assessed several demographic characteristics: gender, education, pension amount, electric bicycle ownership, bus use frequency, and the elderly's psychological effects (perceived physical functioning and psychological factors towards physical activity) in the latent-class logit model. These factors significantly impacted the classification of the latent class at a 10% confidence level. ConclusionThe elderly rely heavily on public transit in China's cities. As such, urban planners should consider the needs of the elderly, and should consider the findings of this study in developing a more comfortable public transit system.

Heterogenous Trip Distance-Based Route Choice Behavior Analysis Using Real-World Large-Scale Taxi Trajectory Data

Most early research on route choice behavior analysis relied on the data collected from the stated preference survey or through small-scale experiments. This manuscript focused on the understanding of commuters’ route choice behavior based on the massive amount of trajectory data collected from occupied taxicabs. The underlying assumption was that travel behavior of occupied taxi drivers can be considered as no different than the well-experienced commuters. To this end, the DBSCAN algorithm and Akaike information criterion (AIC) were first used to classify trips into different categories based on the trip length. Next, a total of 9 explanatory variables were defined to describe the route choice behavior, and and the path size (PS) logit model was then built, which avoided the invalid assumption of independence of irrelevant alternatives (IIA) in the commonly seen multinomial logit (MNL) model. The taxi trajectory data from over 11,000 taxicabs in Xi’an, China, with 40 million trajectory records each day were used in the case study. The results confirmed that commuters’ route choice behavior are heterogenous for trips with varying distances and that considering such heterogeneity in the modeling process would better explain commuters’ route choice behaviors, when compared with the traditional MNL model.

A Mixed Path Size Logit-Based Taxi Customer-Search Model Considering Spatio-Temporal Factors in Route Choice

This paper introduces a model to analyze route choice behavior of taxi drivers for finding next passenger in urban road network. Considering the situation of path overlapping between selected routes in the process of customer-searching, a mixed path size logit (MPSL) model is proposed to analyze route choice behaviors through considering spatio-temporal features of route including customer generation rate, path travel time, cumulative intersection delay, path distance, and path size. Specially, customer generation rate is defined as attraction strength based on historical pick-up records in the route, the intersection travel delay and path travel time are estimated based on large scaled taxi global positioning system (GPS) trajectories. In the experiment, the GPS data were collected from about 36000 taxi vehicles in Beijing at 30-s interval during six months. In the model application, an area of approximately 10 square kilometers in the center of Beijing is selected to demonstrate the effectiveness of the proposed model. The results indicated that the MPSL model could effectively analyze the route choice behavior in customer-searching process and express higher accuracy than traditional multinomial logit model and basic PSL model.

Estimating willingness-to-pay for health care: A discrete choice experiment accounting for non-attendance to the cost attribute.

The literature suggests that discrete choice experiments (DCEs) are a preferable method for estimating willingness-to-pay (WTP) values, as they may avoid some biases often observed in contingent valuation surveys, such as protest and strategic behaviour. However, the choice studies have demonstrated that attribute non-attendance may be a serious issue that leads to biased WTP estimates. A DCE was used to estimate patients' WTP for dental care at the dental school clinic. The endogenous attribute attendance (EAA) model was used to estimate WTP for dental care attributes, while accounting for non-attendance to the cost attribute. The EAA model revealed that almost every second respondent had ignored the cost attribute when making choices, indicating that patients were very sensitive to other characteristics of dental care. Dental care providers should pay particular attention to providing a detailed explanation of treatment to their patients, as this was the most valued attribute of dental care. The welfare estimates from the EAA model, which accounted for non-attendance to the cost attribute, were more than two times lower than welfare estimates from a traditional multinomial logit model and mixed logit model estimated in WTP space. The findings raise concerns for derived WTP estimates in DCE studies that did not consider non-attendance to the cost attribute during the estimation process. Non-attendance to the cost attribute overestimates WTP values, even if being motivated by the true preferences of respondents.

Modeling Stochastic Route Choice Behaviors with Equivalent Impedance

A Logit-based route choice model is proposed to address the overlapping and scaling problems in the traditional multinomial Logit model. The nonoverlapping links are defined as a subnetwork, and its equivalent impedance is explicitly calculated in order to simply network analyzing. The overlapping links are repeatedly merged into subnetworks with Logit-based equivalent travel costs. The choice set at each intersection comprises only the virtual equivalent route without overlapping. In order to capture heterogeneity in perception errors of different sizes of networks, different scale parameters are assigned to subnetworks and they are linked to the topological relationships to avoid estimation burden. The proposed model provides an alternative method to model the stochastic route choice behaviors without the overlapping and scaling problems, and it still maintains the simple and closed-form expression from the MNL model. A link-based loading algorithm based on Dial’s algorithm is proposed to obviate route enumeration and it is suitable to be applied on large-scale networks. Finally a comparison between the proposed model and other route choice models is given by numerical examples.

A tour-based analysis of travel mode choice accounting for regional transit service

The aim of this work is to explore the impact of regional transit service on tour-based commuter travel behavior by using the Bayesian hierarchical multinomial logit model, accounting for the spatial heterogeneity of the people living in the same area. With two indicators, accessibility and connectivity measured at the zone level, the regional transit service is captured and then related to the travel mode choice behavior. The sample data are selected from Washington-Baltimore Household Travel Survey in 2007, including all the trips from home to workplace in morning hours in Baltimore city. Traditional multinomial logit model using Bayesian approach is also estimated. A comparison of the two different models shows that ignoring the spatial context can lead to a misspecification of the effects of the regional transit service on travel behavior. The results reveal that improving transit service at regional level can be effective in reducing auto use for commuters after controlling for socio-demographics and travel-related factors. This work provides insights for interpreting tour-based commuter travel behavior by using recently developed methodological approaches. The results of this work will be helpful for engineers, urban planners, and transit operators to decide the needs to improve regional transit service and spatial location efficiently.

Cross-Nested Joint Model of Travel Mode and Departure Time Choice for Urban Commuting Trips: Case Study in Maryland–Washington, DC Region

The aim of this paper is to contribute to describing the simultaneous choice of travel mode and departure time by making use of a cross-nested logit structure that allows for the joint representation of interalternative correlation along the both choice dimensions. Traditional multinomial logit model and nested logit model are formulated respectively. The analysis uses the revealed preference data collected from Maryland-Washington, DC, regional household travel survey during 2007–2008 for commuting trips, considering more work-related characteristics than previous studies. A comparison of the different model results shows that the presented cross-nested logit structure offers significant improvements over multinomial logit and nested logit models. The empirical results of the analysis reveal significant influences on commuter joint choice behavior of travel mode and departure time. Moreover, a Monte Carlo simulation for two groups of scenarios arising from transportation policies, congestion pricing, and improvements to transit service during peak period is undertaken respectively to examine the impact of a change in car travel cost and transit travel time on the travel mode and departure time switching. The simulation results show that US$5 increase in car travel cost during peak period has a similar effect on reducing drive alone in peak hours as 30% saving in transit travel time but only half of the latter policy in the transit ridership increase.

A joint model of destination and mode choice for urban trips: a disaggregate approach

Trip destination and mode choice are highly influenced by travelers' perceptions and behaviors; selecting a destination and a vehicle for a trip are two interdependent problems. This paper presents and applies a disaggregate joint model for traveler destination and mode choice. The choice model uses fuzzy set and probability theory to deal with the uncertainty embedded in travelers' perceptions and behaviors. The model is structured as a decision tree in which the fuzzy and non-fuzzy classification of influential variables regarding destination selection and mode choice expand the tree. The most influential explanatory variables among all the variables categorized for travelers' household, trip, and living zone specifications are selected based on the maximizing information. An aggregation method is designed to provide aggregate estimates for transportation planning based on the suggested disaggregate choice model. A data-set of over 9000 home-based morning peak-hour trips in Shiraz, a large city in Iran, is used for model construction and evaluation. When compared with a multinomial logit (MNL) model, the suggested models' estimates are more accurate than the traditional MNL model.

Use of Artificial Intelligence for Mode Choice Analysis and Comparison with Traditional Multinomial Logit Model

Travel Demand Forecasting, an essential tool to predict the future demand, is a four stage procedure which involves trip generation, trip distribution, mode choice and traffic assignment, out of which, mode choice analysis plays vital role as it deals with predicting mode used by the travelers to reach their destination. Multinomial Logit (MNL) model is a traditional model adopted for mode choice analysis which has major limitation that the input variables need to have crisp values and hence should be measured accurately which consumes lot of time and resources. Moreover, decision of trip maker for choosing a mode involves human approximations which are not precisely captured by MNL model. This can be overcome by using artificial intelligence techniques like fuzzy logic for modeling mode choice behavior. Fuzzy logic try to harness the human knowledge which is often guided by approximations by accepting input values in linguistic terms. The fuzzy rule base comprises several IF-THEN rules which closely resemble human knowledge and decision-making. In this study, it was thus proposed to apply the concept of fuzzy logic for modeling mode choice and compare the results with traditional MNL model. For this purpose, a total of 5822 samples were collected in Port Blair city, India and data pertaining to input variables viz. in- vehicle travel time, out-vehicle travel time, travel cost and comfort index were considered for development of mode choice models. It was observed that the results obtained from fuzzy logic results gave better prediction accuracy in comparison to the traditional MNL model. Thus it can be concluded that the fuzzy logic models were better able to capture and incorporate the human knowledge and reasoning into mode choice behaviour. Further, developed fuzzy logic models are applied to evaluate selected transport policies to demonstrate the suitability of the developed fuzzy logic mode choice models.

Willingness-to-Pay and Preference Heterogeneity for Rural Bus Attributes

This paper presents estimations of willingness-to-pay (WTP) values in the context of rural bus service in a developing country. Using stated choice data collected from rural bus users, WTP values are estimated by developing standard multinomial logit model (MNL) and three different random parameter logit (RPL) models: One with independent or uncorrelated responses, another taking into account the correlations among responses, and the other allowing heterogeneity around the mean of in-vehicle travel time. While developing the RPL models, successful application of the sparsely used constrained triangular distribution is demonstrated. WTP values are estimated separately for commuting and noncommuting trips. Higher WTP values for travel time are observed for commuting trips, while for noncommuting trips higher WTP values are found for qualitative attribute. Preference heterogeneity associated with the mean is investigated, and the “travel distance” is found to have a statistically significant decomposition effect on the mean of in-vehicle travel time for commuting trips. The traditional MNL model, in general, is found to underestimate the WTP values as compared to RPL models.

Generalized logit model

The logit model is the simplest and best-known probabilistic choice model. Nevertheless according to the deficient flexibility there are problems of making use of the multinomial logit model. In this paper a generalized logit model, which is essentially more flexible than the traditional multinomial logit model, is presented. Furthermore, the usual modal choice models are compared in respect to their flexibility. This is done by calculating the partial derivatives of the choice probability functions at a fixed point. The generalized logit model shows the same flexibility (in a precise sense) as the probit model, but is much more tractable. At the end of this paper an example is set to show the differences in the functional forms between the new model and the traditional models (logit and probit).