-
A Multi-day Needs-based Modeling Approach for Activity and Travel Demand Analysis
Authors:
Kexin Chen,
**** Guan,
Ravi Seshadri,
Varun Pattabhiraman,
Youssef Medhat Aboutaleb,
Ali Shamshiripour,
Chen Liang,
Xiaochun Zhang,
Moshe Ben-Akiva
Abstract:
This paper proposes a multi-day needs-based model for activity and travel demand analysis. The model captures the multi-day dynamics in activity generation, which enables the modeling of activities with increased flexibility in time and space (e.g., e-commerce and remote working). As an enhancement to activity-based models, the proposed model captures the underlying decision-making process of acti…
▽ More
This paper proposes a multi-day needs-based model for activity and travel demand analysis. The model captures the multi-day dynamics in activity generation, which enables the modeling of activities with increased flexibility in time and space (e.g., e-commerce and remote working). As an enhancement to activity-based models, the proposed model captures the underlying decision-making process of activity generation by accounting for psychological needs as the drivers of activities. The level of need satisfaction is modeled as a psychological inventory, whose utility is optimized via decisions on activity participation, location, and duration. The utility includes both the benefit in the inventory gained and the cost in time, monetary expense as well as maintenance of safety stock. The model includes two sub-models, a Deterministic Model that optimizes the utility of the inventory, and an Empirical Model that accounts for heterogeneity and stochasticity. Numerical experiments are conducted to demonstrate model scalability. A maximum likelihood estimator is proposed, the properties of the log-likelihood function are examined and the recovery of true parameters is tested. This research contributes to the literature on transportation demand models in the following three aspects. First, it is arguably better grounded in psychological theory than traditional models and allows the generation of activity patterns to be policy-sensitive (while avoiding the need for ad hoc utility definitions). Second, it contributes to the development of needs-based models with a non-myopic approach to model multi-day activity patterns. Third, it proposes a tractable model formulation via problem reformulation and computational enhancements, which allows for maximum likelihood parameter estimation.
△ Less
Submitted 23 December, 2023;
originally announced December 2023.
-
Discrete Choice Analysis with Machine Learning Capabilities
Authors:
Youssef M. Aboutaleb,
Mazen Danaf,
Yifei Xie,
Moshe Ben-Akiva
Abstract:
This paper discusses capabilities that are essential to models applied in policy analysis settings and the limitations of direct applications of off-the-shelf machine learning methodologies to such settings. Traditional econometric methodologies for building discrete choice models for policy analysis involve combining data with modeling assumptions guided by subject-matter considerations. Such con…
▽ More
This paper discusses capabilities that are essential to models applied in policy analysis settings and the limitations of direct applications of off-the-shelf machine learning methodologies to such settings. Traditional econometric methodologies for building discrete choice models for policy analysis involve combining data with modeling assumptions guided by subject-matter considerations. Such considerations are typically most useful in specifying the systematic component of random utility discrete choice models but are typically of limited aid in determining the form of the random component. We identify an area where machine learning paradigms can be leveraged, namely in specifying and systematically selecting the best specification of the random component of the utility equations. We review two recent novel applications where mixed-integer optimization and cross-validation are used to algorithmically select optimal specifications for the random utility components of nested logit and logit mixture models subject to interpretability constraints.
△ Less
Submitted 21 January, 2021;
originally announced January 2021.
-
Learning Structure in Nested Logit Models
Authors:
Youssef M. Aboutaleb,
Moshe Ben-Akiva,
Patrick Jaillet
Abstract:
This paper introduces a new data-driven methodology for nested logit structure discovery. Nested logit models allow the modeling of positive correlations between the error terms of the utility specifications of the different alternatives in a discrete choice scenario through the specification of a nesting structure. Current nested logit model estimation practices require an a priori specification…
▽ More
This paper introduces a new data-driven methodology for nested logit structure discovery. Nested logit models allow the modeling of positive correlations between the error terms of the utility specifications of the different alternatives in a discrete choice scenario through the specification of a nesting structure. Current nested logit model estimation practices require an a priori specification of a nesting structure by the modeler. In this we work we optimize over all possible specifications of the nested logit model that are consistent with rational utility maximization. We formulate the problem of learning an optimal nesting structure from the data as a mixed integer nonlinear programming (MINLP) optimization problem and solve it using a variant of the linear outer approximation algorithm. We exploit the tree structure of the problem and utilize the latest advances in integer optimization to bring practical tractability to the optimization problem we introduce. We demonstrate the ability of our algorithm to correctly recover the true nesting structure from synthetic data in a Monte Carlo experiment. In an empirical illustration using a stated preference survey on modes of transportation in the U.S. state of Massachusetts, we use our algorithm to obtain an optimal nesting tree representing the correlations between the unobserved effects of the different travel mode choices. We provide our implementation as a customizable and open-source code base written in the Julia programming language.
△ Less
Submitted 18 August, 2020;
originally announced August 2020.
-
Sparse Covariance Estimation in Logit Mixture Models
Authors:
Youssef M Aboutaleb,
Mazen Danaf,
Yifei Xie,
Moshe Ben-Akiva
Abstract:
This paper introduces a new data-driven methodology for estimating sparse covariance matrices of the random coefficients in logit mixture models. Researchers typically specify covariance matrices in logit mixture models under one of two extreme assumptions: either an unrestricted full covariance matrix (allowing correlations between all random coefficients), or a restricted diagonal matrix (allowi…
▽ More
This paper introduces a new data-driven methodology for estimating sparse covariance matrices of the random coefficients in logit mixture models. Researchers typically specify covariance matrices in logit mixture models under one of two extreme assumptions: either an unrestricted full covariance matrix (allowing correlations between all random coefficients), or a restricted diagonal matrix (allowing no correlations at all). Our objective is to find optimal subsets of correlated coefficients for which we estimate covariances. We propose a new estimator, called MISC, that uses a mixed-integer optimization (MIO) program to find an optimal block diagonal structure specification for the covariance matrix, corresponding to subsets of correlated coefficients, for any desired sparsity level using Markov Chain Monte Carlo (MCMC) posterior draws from the unrestricted full covariance matrix. The optimal sparsity level of the covariance matrix is determined using out-of-sample validation. We demonstrate the ability of MISC to correctly recover the true covariance structure from synthetic data. In an empirical illustration using a stated preference survey on modes of transportation, we use MISC to obtain a sparse covariance matrix indicating how preferences for attributes are related to one another.
△ Less
Submitted 14 January, 2020;
originally announced January 2020.
