Esta dissertação teve como objetivo principal investigar a influência, no comportamento de escolha de linha de passageiros de ônibus, de painéis eletrônicos em pontos de parada com informação em tempo real sobre previsões para passagem dos veículos. Para a coleta de dados, um experimento de escolha declarada com desenho eficiente foi aplicado a uma amostra de 1179 entrevistados residentes na Região Metropolitana de São Paulo (RMSP) e pertencentes à comunidade da Universidade de São Paulo. Em cada situação de escolha foram apresentadas duas linhas diferentes de ônibus que chegariam ao mesmo destino, mas não passavam pelo mesmo ponto, de modo que a escolha a ser feita era entre uma ou outra combinação de linha e ponto. Adicionalmente, apenas um dos pontos tinha painel eletrônico. As alternativas também se diferenciavam pelos seguintes atributos: intervalo de tempo programado para a passagem de veículos consecutivos da mesma linha; possível atraso em relação ao intervalo programado; tempo de viagem dentro do veículo até o destino; ocupação do veículo quando chega ao ponto; e valor da tarifa. Na amostra predominaram jovens com até 25 anos (64% da amostra), homens (60%), usuários frequentes de ônibus (80%), estudantes (81%) e entrevistados com pelo menos um automóvel no domicílio (76%). A partir das respostas ao experimento foram estimados modelos de escolha discreta Mixed Logit Panel, de modo a mensurar a importância relativa de cada atributo na decisão e também medir a variabilidade das preferências entre os entrevistados. Os resultados indicam que a presença do painel no ponto de ônibus tem, sim, influência sobre a escolha da linha. Os entrevistados estariam dispostos a pagar em média, pela presença de painel, R$0,12 adicionais, equivalentes a 5 minutos de viagem. Verificou-se também que a existência de painel no ponto diminui a desutilidade marginal da espera, e isto ocorre com mais intensidade quando ela está associada ao atraso do que quando está associada ao intervalo programado entre veículos. O valor médio do tempo de viagem foi relativamente baixo: R$1,44/hora, provavelmente devido à composição socioeconômica da amostra, com elevada proporção de estudantes. No entanto, observou-se que o comportamento de escolha de linha é bastante afetado pelas características socioeconômicas e de uso de ônibus dos entrevistados, podendo o valor do tempo chegar a R$17,00/hora, e a disposição a pagar pelo painel a R$0,77. Acredita-se que os resultados desta pesquisa permitem ampliar o entendimento do comportamento de escolha de linha, ao incorporar a presença de painel no ponto como elemento adicional da decisão. / The main objective of this research is to investigate the influence on bus route choice behavior of variable message signs (VMS) displaying real time predictions of bus arrival at stops. A stated choice survey was conducted, using an efficient design experiment. Sampled individuals were asked to answer to eight choice situations, each presenting two bus routes going to the same destination but with different itineraries and boarding stops. The choice was made between two combinations of bus route and boarding stop; only one of the stops had VMS. The other attributes characterizing alternatives were: bus route headway, (possible) delay at arriving at the stop, travel time until destination, level of vehicle crowdedness when arriving at the boarding stop, and fare. Data were collected from 1179 individuals, mostly students, professors and employees of the University of São Paulo, and all of them residents of the São Paulo Metropolitan Area. The typical interviewee was 25 years old or younger (64% of the sample), male (60%), a frequent bus user (80%), student (81%) and had at least one car in his household. Mixed logit panel discrete choice models were estimated to analyze the data, capturing both the relative importance of each attribute in the decision process and systematic taste variation among individuals. Results show that VMS displaying predictions of bus arrival at stops do influence bus route choice behavior. The estimated average willingness to pay for a bus stop to have a VMS was R$0.12, which corresponds to 5 minutes of travel time. It was also observed that the marginal disutility of waiting time decreases when there is a VMS at the stop. Disutility of waiting due to delays also decreases (more intensively) with the VMS. The average value of travel time was relatively low, compared to expectations: R$1.44/hour, probably due to the socioeconomic profile of the sample, particularly the high proportion of students. Nevertheless, frequency of bus use and socioeconomic characteristics significantly affect route choice behavior; the value of travel time, for instance, may reach R$17/hour, while willingness to pay for a VMS in a stop may become R$0.77. The results indicate that incorporating the VMS as an additional component of the decision, allows for a better understanding of bus route choice behavior.

This research evaluates data-driven demand models using natural language processing techniques and analytical approaches. The first essay offers a comprehensive review of data-driven newsvendor literature and applies natural language processing techniques, including latent semantic analysis, latent Dirichlet allocation and cluster analysis to analyze the text data. This study highlights emerging trends and future research directions in the field of data-driven newsvendor research. The second essay contributes to the data-driven newsvendor inventory management literature by proposing nonparametric approaches that include Tobit and quantile regression incorporating leverage values under conditions of homogeneity and heterogeneity. Lastly, the third essay addresses the optimization of healthcare facility location and resource allocation in post-earthquake scenarios, presenting a linear programming model with telemedicine integration for effective disaster response. This study applies the model to the 2005 Kashmir earthquake in Pakistan. These essays collectively highlight the potential of data-driven methodologies in enhancing decision-making processes across diverse domains, while also pointing towards future research directions to address inherent complexities and uncertainties of the models.

This dissertation provides a comprehensive examination of the role of Transportation Networking Companies (TNC) across four dimensions. First, we examined the factors affecting Transportation Networking Companies (TNC) pricing and destination choice behavior using weekday TNC trip data from Chicago spanning January 2019 through December 2019. Towards achieving this goal, we developed a joint model framework where trip fare is modelled using linear regression model (LR) and destination choice is modelled using a multinomial logit model (MNL). Second, we build a systematic framework to analyze spatial TNC demand patterns (origins) across the urban region at the census tract (CT) level and compare them to overall transportation demand. We propose and compute a novel metric at the census tract level to identify the potential imbalance between overall transportation demand and TNC demand by developing a Generalized Ordered Logit. The model applicability is further illustrated through elasticity analysis. Third, based on earlier studies we identified that current TNC related macroscopic studies do not incorporate attributes at the microscopic resolution. We bridge the macroscopic and microscopic analysis using a bi-level modeling approaching that accommodates for the influence of microscopic attributes within the macroscopic modeling approach. In this proposed framework, the trip level destination choice model (microscopic model) takes the form of a multinomial logit model and the origin-destination flow model (macroscopic model) takes the form of a multinomial logit fractional split model. Finally, in our effort to incorporate TNCs into travel demand tools, we conducted a comprehensive literature review on studies examining the impact of TNCs on various components of travel demand models (TDM). We provide guidelines for potential travel demand model updates using three use case examples including vehicle ownership model, trip generation model, and trip level mode choice components.

A identificação dos fatores que influenciam o processo de escolha do tomador na demanda por crédito apresenta não apenas um interesse mercadológico, mas também em termos acadêmicos e para o formulador de políticas públicas, buscando determinar os impactos de uma decisão que influencia o ambiente macroeconômico, bem como o comportamento dos agentes. Nestes termos, os modelos tradicionais de análise da demanda muitas vezes apresentam suposições pouco realistas e bastante restritivas, necessárias para o processo de estimação dos parâmetros de interesse. Os Modelos Lineares Generalizados Mistos com Variáveis Latentes (GLLAMM) constituem uma classe de modelos que abrangem os tradicionais modelos lineares generalizados e os modelos lineares generalizados mistos, possibilitando uma maior flexibilidade na combinação de um processo de escolha discreta com a determinação dos valores demandados de forma contínua, não impondo um processo único para todas as instituições analisadas, nem tão pouco do processo de escolha. Esta classe de modelos foi aplicada para estudar a demanda por empréstimos bancários utilizando-se informações de uma rica base de dados, a Central de Risco de Crédito do Banco Central. Assim, foi possível a identificação de variáveis como a duração da operação e a classificação de risco da operação que apresentam uma maior relevância no processo de escolha do banco, enquanto que outras, como as garantias, mostraram-se mais importante no volume a ser demandado. A identificação de nichos específicos de algumas instituições foi possível a partir desta análise. A flexibilidade desta classe de modelos também foi utilizada no intuito de se identificar os fatores que influenciam a demanda por crédito pelos bancos nas operações de redesconto, tendo conseguido tratar o problema de superdispersão ocasionado pelo excesso de zeros neste conjunto de dados. Adicionalmente, tanto efeitos diretos quanto indiretos da taxa de redesconto foram possíveis de serem estudados a partir da inclusão de efeitos aleatórios tanto no intercepto, possibilitando a incorporação de efeitos específicos de cada instituição financeira, bem como nos coeficientes, captando comportamentos individuais de cada banco frente a um mesmo estímulo. / The aim of this study is to identify the variables that affect borrower´s decision-making process through the estimation of loan demand equations. This research is relevant not only for market practitioners, but also to academics and to policy-makers, concerned with the evaluation of possible impacts of decision on the economic environment and on the agent´s behavior. Traditional models for demand estimation make unreasonable and very restrictive assumptions to estimate the parameters of interest. Generalized Linear Latent and Mixed Models (GLLAMM) constitute a class of models that includes the traditional Generalized Linear Models (GLM) and Generalized Linear Mixed Models (GLMM), which offer more flexibility and they are particularly suitable to situations that combine discrete choice with continuous decisions. Therefore, we can estimate individual equation for each bank simultaneously. This class of models was applied to study the demand for bank loans using a rich dataset provided by the Central Bank?s Credit Risk Bureau. Among the analyzed variables, the loan maturity and the credit risk classification were important in the bank choice while warranties were important in the decision related to the amount borrowed. Moreover, we could also identify specific market segments for some banks. The flexibility of this class of models was also used to identify the factors affecting the demand for overnight funds by commercial banks. This model overcomes overdispersion problems caused by excess of zeros found in the dataset. Additionally, we identified the direct and indirect effects of rediscount rate through the inclusion of random effects in the intercept (incorporating specific effects for each bank) and other coefficients (identifying individual behavior of each bank regarding the same stimuli).

Spatial and social dependence shape human activity-travel pattern decisions and their antecedent choices. Although the transportation literature has long recognized the importance of considering spatial and social dependencies in modeling individuals’ choice behavior, there has been less research on techniques to accommodate these dependencies in discrete choice models, mainly because of the modeling complexities introduced by such interdependencies. The main goal of this dissertation, therefore, is to propose new modeling approaches for accommodating flexible spatial and social dependency structures in discrete choice models within the broader context of activity-based travel demand modeling. The primary objectives of this dissertation research are three-fold. The first objective is to develop a discrete choice modeling methodology that explicitly incorporates spatial dependency (or correlation) across location choice alternatives (whether the choice alternatives are contiguous or non-contiguous). This is achieved by incorporating flexible spatial correlations and patterns using a closed-form Generalized Extreme Value (GEV) structure. The second objective is to propose new approaches to accommodate spatial dependency (or correlation) across observational units for different aspatial discrete choice models, including binary choice and ordered-response choice models. This is achieved by adopting different copula-based methodologies, which offer flexible dependency structures to test for different forms of dependencies. Further, simple and practical approaches are proposed, obviating the need for any kind of simulation machinery and methods for estimation. Finally, the third objective is to formulate an enhanced methodology to capture the social dependency (or correlation) across observational units. In particular, a clustered copula-based approach is formulated to recognize the potential dependence due to cluster effects (such as family-related effects) in an ordered-response context. The proposed approaches are empirically applied in the context of both spatial and aspatial choice situations, including residential location and activity participation choices. In particular, the results show that ignoring spatial and social dependencies, when present, can lead to inconsistent and inefficient parameter estimates that, in turn, can result in misinformed policy actions and recommendations. The approaches proposed in this research are simple, flexible and easy-to-implement, applicable to data sets of any size, do not require any simulation machinery, and do not impose any restrictive assumptions on the dependency structure. / text

This thesis presents a prism- and gap-based approach for modelling shopping destination choice in the Travel/Activity Scheduler for Household Agents (TASHA). The gap-location choice model improves upon TASHA’s existing destination choice model in 3 key ways: 1) Shifting from a zone-based to a disaggregate location choice model, 2) Categorizing shopping trips into meaningful types, and 3) Accounting for scheduling constraints in choice set generation and location choice. The model replicates gap and location choices reasonably well at an aggregate level and shows that a simple yet robust model can be developed with minimal changes to TASHA’s existing location choice model. The gap-based approach to destination choice is envisioned as a small but significant step towards a more comprehensive location choice model in a dynamic activity scheduling environment.

This thesis presents a prism- and gap-based approach for modelling shopping destination choice in the Travel/Activity Scheduler for Household Agents (TASHA). The gap-location choice model improves upon TASHA’s existing destination choice model in 3 key ways: 1) Shifting from a zone-based to a disaggregate location choice model, 2) Categorizing shopping trips into meaningful types, and 3) Accounting for scheduling constraints in choice set generation and location choice. The model replicates gap and location choices reasonably well at an aggregate level and shows that a simple yet robust model can be developed with minimal changes to TASHA’s existing location choice model. The gap-based approach to destination choice is envisioned as a small but significant step towards a more comprehensive location choice model in a dynamic activity scheduling environment.

[pt] A análise de demanda condicional (conditional demand analysis - CDA) é um método econométrico que, aplicado ao estudo do consumo residencial de energia elétrica, permite estimar a quantidade de energia consumida por diferentes aparelhos eletrodomésticos. Nessa tese, métodos bayesianos são utilizados na estimação dos modelos CDA. A restrição de não negatividade dos coeficientes de consumo de energia é incorporada ao modelo através do uso da densidade normal truncada como priori dos parâmetros. Como as densidades a posteriori resultantes também são truncadas, métodos de simulação estocástica cria cadeias de Markov são usados na estimação de tais densidades. O método desenvolvido é aplicado a um conjunto de dados fornecido pela LIGHT, uma das concessionárias de energia do estado do Rio de Janeiro, gerando as curvas de carga para diversos aparelhos. / [en] Conditional demand analysis (CDA) is an econometric method that, applied to studies of consumption of energy in the household sector, allows us to estimate the demand of energy for different appliances. In this thesis, the estimation of the CDA models is made in a Bayesian framework. The truncated normal distribution is used as a prior of the parameters, assuring their nonnegativity restrictions. Since the resulting posteriors are truncated distributions too, the Gibbs sampler is applied in the estimation of those densities. The results obtained are applied to a dataset obtained from LIGHT, one of the electricity utilities of Rio de Janeiro, in order to obtain some appliances load curves.

A dissertação traz um breve resumo da história da siderurgia no Brasil, e comentários sobre o que vem ocorrendo no setor no mundo. A dissertação busca contribuir com a estimação de demanda para mercados oligopolizados, utilizando como proxie o mercado siderúrgico brasileiro. O objetivo, portanto, é estimar as variáveis econômicas que impactam o consumo de vergalhão no Brasil. Para isso são criados diversos modelos econométricos de demanda, utilizando as modelagens VAR (Vetor Auto-regressivo), BVAR (Vetor Autoregressivo Bayesiano) e Variáveis Instrumentais (IV). A metodologia BVAR foi aquela que apresentou os melhores resultados, com os seus coeficientes sendo robustos e estatisticamente significantes, além de reproduzirem a teoria econômica. / The present dissertation begins with a small resume of the history of the steel sector in Brazil, added to this some comments about that in the world. The dissertation aims to contribute with the estimation of oligopoly markets demand, using as a proxy the steel industry in Brazil. The main objective is to estimate the economics variables witch impact the rebar consumption in Brazil. The estimates are made using different econometric methodologies as VAR (Vector Autoregressive), BVAR (Bayesian Vector Autoregressive) and Instrumental Variables. The BVAR methodology is the one witch presents the best results, with the coefficients signs being robust and statistic significant, besides reproducing the economic theory.

A dissertação traz um breve resumo da história da siderurgia no Brasil, e comentários sobre o que vem ocorrendo no setor no mundo. A dissertação busca contribuir com a estimação de demanda para mercados oligopolizados, utilizando como proxie o mercado siderúrgico brasileiro. O objetivo, portanto, é estimar as variáveis econômicas que impactam o consumo de vergalhão no Brasil. Para isso são criados diversos modelos econométricos de demanda, utilizando as modelagens VAR (Vetor Auto-regressivo), BVAR (Vetor Autoregressivo Bayesiano) e Variáveis Instrumentais (IV). A metodologia BVAR foi aquela que apresentou os melhores resultados, com os seus coeficientes sendo robustos e estatisticamente significantes, além de reproduzirem a teoria econômica. / The present dissertation begins with a small resume of the history of the steel sector in Brazil, added to this some comments about that in the world. The dissertation aims to contribute with the estimation of oligopoly markets demand, using as a proxy the steel industry in Brazil. The main objective is to estimate the economics variables witch impact the rebar consumption in Brazil. The estimates are made using different econometric methodologies as VAR (Vector Autoregressive), BVAR (Bayesian Vector Autoregressive) and Instrumental Variables. The BVAR methodology is the one witch presents the best results, with the coefficients signs being robust and statistic significant, besides reproducing the economic theory.