Tag Archive: Choudhury

 Car Trip Generation Models in the Developing World: Data Issues and Spatial Transferability.

In many countries of the developing world, it is difficult to conduct large-scale household travel surveys to collect data for travel behaviour model estimation and application. This paper focuses on two candidate solutions to the problem: (1) developing models that can be applied for prediction using secondary data collected for other purposes and include socio-demographic information but do not include transport specific information such as the car and/or transit pass ownership (e.g. census, public health records, etc.), (2) ‘borrowing’ a model developed using data from a similar city within the same region. In the first approach, we investigate the feasibility of developing car trip generation models which imputes the car ownership variable with estimated car ownership propensities. The proposed framework is applied in two East African cities, Nairobi and Dar-es-Salaam. The estimation results indicate that for both cities the proposed approach outperforms the models that exclude the car ownership variable. In the second approach, we investigate the spatial transferability of the models developed in the first approach between the two cities to evaluate if it is justified to apply models from one developing country to another in the absence of local models. Results indicate that though some of the estimated parameters are not significantly different from each other between the two cities, statistical tests do not support direct transferability of all the models from Nairobi to Dar-es-Salaam or vice versa. However, interestingly, the simpler model (which excludes car-ownership) outperforms the model with imputed car ownership propensity in terms of transferability. These findings provide useful insights into the development of trip generation models under data constraints which can practically be very useful for developing countries.

Bwambale, A., Choudhury, C. F., & Sanko, N. (2019). Car Trip Generation Models in the Developing World: Data Issues and Spatial Transferability. Transportation in Developing Economies, 5(2), 10.


Modelling residential mobility decision and its impact on car ownership and travel mode.

Household residential relocation can happen at different scales – local, regional national and international. The impacts of the different scales of residential relocation is likely to have varying impacts on mid-term (e.g. car or transit pass ownership) and day-to-day mobility decisions (e.g. mode choice for a specific trip for example). These mobility changes can be of different levels as well. For example, there are differences between the decision to transition from owning no car to one car and from one car to two cars. Identifying which factors affect the different magnitudes of mobility changes and quantifying the impact of various scales of residential relocation on these changes are crucial to better understanding of travel behaviour. The present study uses discrete choice models on revealed preference data to address these research questions. To complement the travel behaviour models, a residential relocation model has also been developed to predict the probability of a household to stay in the current location vs. to move locally, regionally or nationally at a given point of time. Given that the residential relocations are rare events, the British household panel survey (BHPS) spanning 18 years has been used to model the choices made by the same households in terms of residential relocation, car ownership and commute mode of the household head. Our results indicate that sociodemographic characteristics, travel behaviour and life events of the households have a significant effect on relocation, car ownership and commute mode choice. As expected, the parameters of the car ownership and commute mode choice models vary significantly with the type of relocation. Further, the socio-demographic factors and life-events also have a varying impact on the scale of relocation. The residential relocation, car ownership and commute mode choice models developed in this research can be used to better predict the medium and long term changes in travel behaviour over course of time.

Haque, Md B., Choudhury, C.F., Hess, S. & Crastes dit Sourd, R. (2019), Modelling residential mobility decision and its impact on car ownership and travel mode. Travel Behaviour and Society, Volume 17, October 2019, Pages 104-119.

Combining driving simulator and physiological sensor data in a latent variable model to incorporate the effect of stress in car-following behaviour.

Car-following models, which are used to predict the acceleration-deceleration decisions of drivers in the presence of a closely spaced lead vehicle, are critical components of traffic microsimulation tools and useful for safety evaluation. Existing car-following models primarily account for the effects of surrounding traffic conditions on a driver’s decision to accelerate or decelerate. However, research in human factors and safety has demonstrated that driving decisions are also significantly affected by individuals’ characteristics and their emotional states like stress, fatigue, etc. This motivates us to develop a car-following model where we explicitly account for the stress level of the driver and quantify its impact on acceleration-deceleration decisions. An extension of the GM stimulus-response model framework is proposed in this regard, where stress is treated as a latent (unobserved) variable, while the specification also accounts for the effects of drivers’ sociodemographic characteristics. The proposed hybrid models are calibrated using data collected with the University of Leeds Driving Simulator where participants are deliberately subjected to stress in the form of aggressive surrounding vehicles, slow leaders and/or time pressure while driving in a motorway setting. Alongside commonly used variables, physiological measures of stress (i.e. heart rate, blood volume pulse, skin conductance) are collected with a non-intrusive wristband. These measurements are used as indicators of the latent stress level in a hybrid model framework and the model parameters are estimated using Maximum Likelihood Technique. Estimation results indicate that car-following behaviour is significantly influenced by stress alongside speed, headway and drivers’ characteristics. The findings can be used to improve the fidelity of simulation tools and designing interventions to improve safety.

Paschalidis, E., Choudhury, C.F. & Hess, S. (2019), Combining driving simulator and physiological sensor data in a latent variable model to incorporate the effect of stress in car-following behaviour. Analytic Methods in Accident Research, 22, 100089.

Modelling Residential Location Choices with Implicit Availability of Alternatives.

Choice set generation is a challenging aspect of disaggregate level residential location choice modelling due to the large number of candidate alternatives in the universal choice set (hundreds to hundreds of thousands). The classical Manski method (Manski, 1977) is infeasible here because of the explosion of the number of possible choice sets with the increase in the number of alternatives. Several alternative approaches have been proposed in recent years to deal with this issue, but these have limitations alongside strengths. For example, the Constrained Multinomial Logit (CMNL) model (Martínez et al., 2009) offers gains in efficiency and improvements in model fit but has weaknesses in terms of replicating the Manski model parameters. The rth-order Constrained Multinomial Logit (rCMNL) model (Paleti, 2015) performs better than the CMNL model in producing results consistent with the Manski model, but the benefits disappear when the number of alternatives in the universal choice set increases. In this study, we propose an improved CMNL model (referred to as Improved Constrained Multinomial Logit Model, ICMNL) with a higher order formulation of the CMNL penalty term that does not depend on the number of alternatives in the choice set. Therefore, it is expected to result in better model fit compared to the CMNL and the rCMNL model in cases with large universal choice sets. The performance of the ICMNL model against the CMNL and the rCMNL model is evaluated in an empirical study of residential location choices of households living in the Greater London Area. Zone level models are estimated for residential ownership and renting decisions where the number of alternatives in the universal choice set is 498 in each case. The performance of the models is examined both on the estimation sample and the holdout sample used for validation. The results of both ownership and renting models indicate that the ICMNL model performs considerably better compared to the CMNL and the rCMNL model for both the estimation and validation samples. The ICMNL model can thus help transport and urban planners in developing better prediction tools.

Haque, M.D., Choudhury, C.F. & Hess, S. (2019), Modelling Residential Location Choices with Implicit Availability of Alternatives. Journal of Transport and Land Use, 12(1), pp. 597–618.

New appraisal values of travel time saving and reliability in Great Britain.

This paper provides an overview of the study ‘Provision of market research for value of time savings and reliability’ undertaken by the Arup/ITS Leeds/Accent consortium for the UK Department for Transport (DfT). The paper summarises recommendations for revised national average values of in-vehicle travel time savings, reliability and time-related quality (e.g. crowding and congestion), which were developed using willingness-to-pay (WTP) methods, for a range of modes, and covering both business and non-work travel purposes. The paper examines variation in these values by characteristics of the traveller and trip, and offers insights into the uncertainties around the values, especially through the calculation of confidence intervals. With regards to non-work, our recommendations entail an increase of around 50% in values for commute, but a reduction of around 25% for other non-work—relative to previous DfT ‘WebTAG’ guidance. With regards to business, our recommendations are based on WTP, and thus represent a methodological shift away from the cost saving approach (CSA) traditionally used in WebTAG. These WTP-based business values show marked variation by distance; for trips of less than 20 miles, values are around 75% lower than previous WebTAG values; for trips of around 100 miles, WTP-based values are comparable to previous WebTAG; and for longer trips still, WTP-based values exceed those previously in WebTAG.

Batley, R.P., Bates, J., Bliemer, M., Börjesson, M., Bourdon, J., Ojeda Cabral, M., Chintakayala, P.K., Choudhury, C., Daly, A.J., Dekker, T., Drivyla, E., Fowkes, A., Hess, S., Heywood, C., Johnson, D., Laird, J., Mackie, P., Parkin, J., Sanders, S., Sheldon, R., Wardman, M. & Worsley, T. (2019), New appraisal values of travel time saving and reliability in Great Britain. Transportation, 46, Issue 3, pp 583–621.

Modelling long-distance route choice using mobile phone call detail record data: A case study of Senegal.

The growing mobile phone penetration rates have led to the emergence of large-scale call detail records (CDRs) that could serve as a low-cost data source for travel behaviour modelling. However, to the best of our knowledge, there is no previous study evaluating the potential of CDR data in the context of route choice behaviour modelling. Being event-driven, the data are discontinuous and only able to yield partial trajectories, thus presenting serious challenges for route identification. This paper proposes techniques for inferring the users’ chosen routes or subsets of their likely routes from partial CDR trajectories, as well as data fusion with external sources of information such as route costs, and then adapts the broad choice framework to the current modelling scenario. The model results show that CDR data can capture the expected travel behaviour and the derived values of travel time are found to be realistic for the study area.

Bwambale, A., Choudhury, C.F. & Hess, S. (2019), Modelling long-distance route choice using mobile phone call detail record data: A case study of Senegal. Transportmetrica A, 15(2), pp. 1543-1568.

Mode choice with latent availability and consideration: theory and a case study.

Over the last two decades, passively collected data sources, like Global Positioning System (GPS) traces from data loggers and smartphones, have emerged as a very promising source for understanding travel behaviour. Most choice model applications in this context have made use of data collected specifically for choice modelling, which often has high costs associated with it. On the other hand, many other data sources exist in which respondents’ movements are tracked. These data sources have thus far been underexploited for choice modelling. Indeed, although some information on the chosen mode and basic socio-demographic data is collected in such surveys, they (as well as in fact also some purpose collected surveys) lack information on mode availability and consideration. This paper addresses the data challenges by estimating a mode choice model with probabilistic availability and consideration, using a secondary dataset consisting of ‘annotated’ GPS traces. Stated mode availability by part of the sample enabled the specification of an availability component, while the panel nature of the data and explicit incorporation of spatial and environmental factors enabled estimation of latent trip specific consideration sets. The research thus addresses an important behavioural issue (explicit modelling of availability and choice set) in addition to enriching the data for choice modelling purposes. The model produces reasonable results, including meaningful value of travel time (VTT) measures. Our findings further suggest that a better understanding of mode choices can be obtained by looking jointly at availability, consideration and choice.

Calastri, C., Hess, S., Choudhury, C.F., Daly, A.J. & Gabrielli, L. (2019), Mode choice with latent availability and consideration: theory and a case study. Transportation Research Part B, 123, pp. 374-385.

Modelling trip generation using mobile phone data: a latent demographics approach.

Traditional approaches to trip generation modelling rely on household travel surveys which are expensive and prone to reporting errors. On the other hand, mobile phone data, where spatio-temporal trajectories of millions of users are passively recorded has recently emerged as a promising input for transport analyses. However, such data has primarily been used for the development of human mobility models, extraction of statistics on human mobility behaviour, and origin-destination matrix estimation as opposed to the development of econometric models of travel demand. This is primarily due to the exclusion of user demographics from mobile phone data made available for research (owing to privacy reasons). In this study, we address this limitation by proposing a hybrid trip generation model framework where demographic groups are treated as latent or unobserved. The proposed model first predicts the demographic group membership probabilities of individuals based on their phone usage characteristics and then uses these probabilities as weights inside a latent class model for trip generation, with different classes representing different socio-demographic groups. The model is calibrated using the call log data of a sub-sample of users with known demographics and trip rates extracted from their GSM mobility data. The performance of the hybrid model is compared with that of a traditional trip generation model which uses observed demographic variables to validate the proposed methodology. This comparative analysis shows that the model fit and the prediction results of the hybrid model are close to those of the traditional model. The research thus serves as a proof-of-concept that the mobile phone data can be successfully used to develop econometric models of transport planning by having additional information for a subset of the users.

Bwambale, A., Choudhury, C.F. & Hess, S. (2019), Modelling trip generation using mobile phone data: a latent demographics approach. Journal of Transport Geography, 76, Pages 276-286.

Temporal transferability of vehicle ownership models in the developing world: case study of Dhaka, Bangladesh.

Temporal transferability of model parameters is a critical issue, especially in the context of developing countries where data and resources for transport model development are extremely limited. This study investigates the temporal transferability of vehicle ownership models with special emphasis on exploring the effect of model structure on temporal transferability. The performance of potential updating methods for making the models more transferable are also compared. The household survey data collected from Dhaka, Bangladesh in 2005 and 2010 have been used in this regard. Different forms of random utility and count regression models of car, motorcycle, and bicycle ownership have been developed using income and household size, and number of workers, children, and licensed drivers as explanatory variables. The temporal transferability of each model between the two time periods has been compared rigorously using statistical tests. Results indicate that the multinomial logit model has better temporal transferability than the count regression models. In relation to model updating, the combined transfer estimation method for model updating is found to perform better than the Bayesian updating. The findings can provide useful guidance during application of a pre-existing model in the context of a developing country.

Flavia, A., & Choudhury, C. (2019). Temporal transferability of vehicle ownership models in the developing world: case study of Dhaka, Bangladesh. Transportation Research Record, 2673(3), 722-732.

A Tour-Based Mode Choice Model for Commuters in Indonesia.

With the advent of activity-based modelling, transport planners’ focus has shifted from isolated trips to tours. Tours are series of interconnected trips that start and finish at home. There are different types of tours; we focus on two: hwh (start at home; go to work; and then go back home) and hw+wh (where + represents a non-work activity). Tour types introduce a new dimension to the traditional problem of travel mode choice, as the mode choice might be influenced by the type of tour. This study attempts to measure and compare the relationship between tour type and mode choice using three different modelling approaches: Multinomial Logit (MNL); Nested Logit (NL) and Cross-Nested Logit (CNL). We compare each approach using secondary data from a larger survey: 24-h daily activity patterns of 420 commuters between Bekasi and Jakarta; one of the busiest commuting routes in Indonesia. Among other results, we found that gender and income significantly influence commuter’s choice of mode and that reducing travel time and cost can increase the ridership of public transport. Furthermore, the NL and CNL models showed significant improvement over the simpler MNL when grouping the alternatives based on tour types. This points to a significant influence of the tour type on the mode choice. Policy recommendations to increase traveler’s wellbeing are also formulated.

Bastarianto, F. F., Irawan, M. Z., Choudhury, C., Palma, D., & Muthohar, I. (2019). A Tour-Based Mode Choice Model for Commuters in Indonesia. Sustainability, 11(3), 788.

Why live far? — Insights from modeling residential location choice in Bangladesh.

Increasing commute distances often lead to increased auto-dependency and is a major problem in many developed as well as developing countries. While in developed countries, the propensity to commute long distances generally originates from the preference to work in the core of the city and live in the suburb or periphery, in developing countries, the trend is often quite the opposite. For example, in Bangladesh, people generally have a strong preference to live at the heart of the major cities even if they work at the peripheral areas of the city, in another city or in a rural area. Further, it is also not uncommon to maintain split-families where the earning member of the family lives near the workplace while the rest of the family is based in a big city (subject to affordability). These phenomena lead to substantial increase in Vehicle Miles Traveled (VMT) and add burden to the transport infrastructure. The focus of the research is to explore the key factors that induce middle and upper-middle class commuters in Bangladesh to live away from their workplace and/or maintain split-families. A case study is conducted using Stated Preference (SP) surveys conducted among the faculty members of two universities: one located at the periphery of the capital city and the other quite far away. Discrete Choice Models are developed using the collected data. Results reveal that albeit some differences, for both cases, the choices are strongly driven by quality of the education institutes and the house rent. Factors like gender, income and car-ownership, which traditionally play a strong role in the context of developed countries, are found to be of less significance. The models, though estimated with limited data, provide useful insights about the factors that drive residential location choices in the context of a developing country and can help in formulating policies for encouraging people to live closer to their workplaces and thereby reduce commuter VMT.

Choudhury, C.F. & Ayaz, S.B. (2015), Why live far? — Insights from modeling residential location choice in Bangladesh. Journal of Transport Geography, 48, pp 1-9.