Hidden Markov Model-based population synthesis

Saadi, Ismaïl; Mustafà, Ahmed; Teller, Jacques; Farooq, Bilal; Cools, Mario

doi:10.1016/j.trb.2016.04.007

Cited by 65 publications

(38 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, MCMC algorithm has played an important role in travel behaviour modelling problems in transportation, with successful applications in agent-based simulations [22], hybrid choice models [23,24], and population synthesis [25,26,27]. However, in order to match the asymptotic efficiency of maximum likelihood, MCMC draws must grow at a rate faster than the square root of the number of agents [3,28].…”

Section: Literature Reviewmentioning

confidence: 99%

A bi-partite generative model framework for analyzing and simulating large scale multiple discrete-continuous travel behaviour data

Wong

Farooq

2020

Transportation Research Part C: Emerging Technologies

Self Cite

View full text Add to dashboard Cite

The emergence of data-driven demand analysis have led to the increased use of generative modelling to learn the probabilistic dependencies between random variables. Although their apparent use has largely been limited to image recognition and classification in recent years, generative machine learning algorithms can be a powerful tool for travel behaviour research by replicating travel behaviour by the underlying properties of data structures. In this paper, we examine the use of generative machine learning approach for analyzing multiple discrete-continuous (MDC) travel behaviour data. We provide a plausible perspective of how we can exploit the use of machine learning techniques to interpret the underlying heterogeneities in the data. We show that generative models are conceptually similar to choice selection behaviour process through information entropy and variational Bayesian inference. Without loss of generality, we consider a restricted Boltzmann machine (RBM) based algorithm with multiple discrete-continuous layer, formulated as a variational Bayesian inference optimization problem. We systematically describe the proposed machine learning algorithm and develop a process of analyzing travel behaviour data from a generative learning perspective. We show parameter stability from model analysis and simulation tests on an open dataset with multiple discrete-continuous dimensions from a data size of 293,330 observations. For interpretability, we derive the conditional probabilities, elasticities and perform statistical analysis on the latent variables. We show that our model can generate statistically similar data distributions for travel forecasting and prediction and performs better than purely discriminative methods in validation. Our results indicate that latent constructs in generative models can accurately represent the joint distribution consistently on MDC data.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

A bi-partite generative model framework for analyzing and simulating large scale multiple discrete-continuous travel behaviour data

Wong

Farooq

2020

Transportation Research Part C: Emerging Technologies

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many older and also very recent publications originate from the area of passenger transport modelling because it has a longer tradition using a disaggregate modelling paradigm. For example, the work of Saadi et al (2016) proposes using Markov models in this context. Barthelemy and Toint (2013) According to their work, generation procedures can be divided into two categories: Iterative Proportional Fitting (IPF) and Combinatorial Optimisation (CO).…”

Section: Generating Establishment Sizes By Stochastic Simulationmentioning

confidence: 99%

Modelling change in supply-chain-structures and its effect on freight transport demand

Ottemöller

Friedrich

2019

Transportation Research Part E: Logistics and Transportation Re

View full text Add to dashboard Cite

Economic activity, especially in the manufacturing industries, is a major determinant of freight transport. Since the actual manufacturing processes are bound to specialised business establishments, which are only partially dispersed across space, commodity flows are required that connect locations of excess supply with locations that show unfulfilled demand for goods. In this context, economic reasoning of the involved actors leads to the formation of industry specific supply chain structures. Various authors emphasise the interrelation of these continuously evolving supply chain structures and freight transport demand. However, so far no disaggregate model exists that explicitly captures this interdependence.The study at hand addresses this gap by developing a disaggregate model for simulating the impact of change in supply chain structures on the corresponding freight transport demand. The proposed model covers centralisation and vertical disintegration as examples of structural change. For this purpose, the model quantitatively describes spatially disaggregate supply chain structures, consisting of business establishments and commodity flows, on the level of entire sectors. The model development is accompanied by an interdisciplinary literature review that gives an overview of existing research on supply chain structures and freight transport demand.The developed model consists of two phases. A first phase generates an artificial industry landscape of business establishments and commodity flows according to available aggregate statistics. The generation relies on elements of stochastic simulation and directed choice procedures. The model's second phase simulates change in the supply chain structures from the first phase. Using linear programming, a maximum solution range regarding the impact on freight transport is calculated. Increasing the degree of assumptions, the solution space can be narrowed. Here, the model applies a combination of stochastic simulation, linear programming, and fitting procedures.The model is applied for analysing centralisation in the poultry industry and vertical disintegration in the automotive industry of Germany. For both cases, a broad range of data sources is used, e.g. common public statistics on establishment sizes and spatial distribution of employment but also sectoral data, e.g. from industry associations or case studies. The real-world consistency of spatial flow patterns is ensured by assigning commodity flows according to statistical macroscopic flows.Overall, the simulation results show that an increase in freight transport performance is to be expected for the case of centralisation as well as vertical disintegration. However, the maximum solution ranges also indicate that assuming suitable location choice and flow assignment a reduction in freight transport performance is mathematically possible. The analysis also addresses the suitability of state measures for mitigating the impact of changes in the supply chain structure on freight transport demand.In summ...

show abstract

“…They claim it can capture the complex interaction and hierarchical household structure of the sample and show that it is better than other methods (i.e., MCMC, directly inflating (DI) and IPF). Moreover, Saadi, Mustafa, Teller, Farooq, and Cools (2016) used the hidden Markov model (HMM), which is a stochastic model that learns a complex joint distribution sample structure or known as the simplest BN. However, this model has less consideration for hierarchical household structure.…”

Section: Introductionmentioning

confidence: 99%

Integrating Bayesian network and generalized raking for population synthesis in Greater Jakarta

Ilahi

Axhausen

2019

Regional Studies, Regional Science

View full text Add to dashboard Cite

Constructing agent data with detailed information on their sociodemographics is substantially important for agent-based modelling. However, to collect data about the whole population is not efficient, since it requires an expensive and time-consuming survey, especially for a large population. The paper uses a novel approach that integrates Bayesian network (BN) and generalized raking (GR) multilevel iterative proportional fitting (IPF). Furthermore, the approach is applied to construct the population for Greater Jakarta, Indonesia, which consists of 30 million inhabitants. The results show that the BN approach can produce data that represent the probability distribution of sample data and that the IPF can match it against aggregate census data.

show abstract

Hidden Markov Model-based population synthesis

Cited by 65 publications

References 25 publications

A bi-partite generative model framework for analyzing and simulating large scale multiple discrete-continuous travel behaviour data

A bi-partite generative model framework for analyzing and simulating large scale multiple discrete-continuous travel behaviour data

Modelling change in supply-chain-structures and its effect on freight transport demand

Integrating Bayesian network and generalized raking for population synthesis in Greater Jakarta

Contact Info

Product

Resources

About