Recursive Logit Model Research Articles

An empirical analysis of learning and forgetting mechanisms in repeated evacuations: A panel recursive logit approach

Global efforts to address climate change increasingly focus on risk communication to encourage people's mitigation behaviour. This study investigates how risk communication influences evacuation behaviour, through repeated risk communications and stated preference surveys. In modelling the learning process of infrequent events like disasters, it is necessary to consider decision-making under uncertainty about future situations and the tendency to forget acquired information. We propose a recursive model that captures two decision dynamics relevant to infrequent events: proactive decision-making amid dynamically changing conditions and the learning and forgetting processes between events. The former is modelled using the recursive logit model on a time-expanded network, while the latter is represented by dynamic changes in utility and memory decay parameters. The estimation results indicate that risk communication generally promotes earlier evacuation to safer places. However, effective communication strategies and the duration of their effects differ based on individuals' perceived and actual risks.

A dynamic discrete choice modelling approach for forward-looking travel mode choices

In this paper, we present a systematic approach based on dynamic discrete choice models (DDCM) to investigate individuals’ forward-looking mode choice behaviours in daily travel tours with multiple destinations. We propose a novel network transformation model that encompasses the entire decision space of all feasible mode combinations for every observed trip/tour in the dataset. By applying the well-established Recursive Logit model structure commonly used in path choice modelling, we address the tour mode choice problem effectively and quantify forward looking considerations in the mode choice process. The proposed model captures the complex considerations individuals take into account when making mode choices. The network transformation incorporates downstream mode limitations into the preceding mode choice options, enabling us to model individuals’ forward-looking behaviour and gain insights into how considerations of future trips such as a shopping in the evening, or school pick-up trip influence previous mode choice decisions earlier in the day. Uncovering and quantifying these hidden forward-looking factors can help modellers better explain the private car usage usually observed for the entire sequences of daily trips, even in presence of competitive alternative modes. The proposed network transformation also enables us to measure the effect of the requirement/preference to return private vehicles (car, motorcycle, and bicycle) home on mode choices in home-bound trips, and subsequently, on the entire daily mode choice decisions. To validate the proposed model, we utilise the VISTA household travel survey data from the Melbourne Metropolitan area in Australia. The model is compared against baseline models, demonstrating its statistical advantages. Additionally, intuitive illustrations using the data showcase the model’s efficacy. From transport planning and policy perspective, tour-based mode choice modelling provides a more comprehensive and precise understanding of travel behaviour by considering the sequence of trips made by an individual. This can help capture the interactions and dependencies between different trips, which trip-based models may overlook. The proposed model is more suitable for analysing the effects of policy interventions like congestion pricing, public transport investments, or new mobility initiatives, as they can better represent the cascading effects of such policies across multiple trips.

Recursive logit-based meta-inverse reinforcement learning for driver-preferred route planning

Driver-preferred route planning often evaluates the quality of a planned route based on how closely it is followed by the driver. Despite decades of research in this area, there still exist nonnegligible deviations from planned routes. Recently, with the prevalence of GPS data, Inverse Reinforcement Learning (IRL) has attracted much interest due to its ability to directly learn routing patterns from GPS trajectories. However, existing IRL methods are limited in that: (1) They rely on numerical approximations to calculate the expected state visitation frequencies (SVFs), which are inaccurate and time-consuming; and (2) They ignore the fact that the coverage of GPS trajectories is skewed toward popular road segments, causing difficulties in learning from sparsely covered ones. To overcome these challenges, we propose a recursive logit-based meta-IRL approach, where (1) We use the recursive logit model to capture drivers’ route choice behavior so that the expected SVFs can be analytically derived, which substantially reduces the computational efforts; and (2) We introduce meta-parameters and employ meta-learning techniques so that the learning on sparsely covered road segments can benefit from that on popular ones. When training our IRL model, we update the rewards of road segments with the expected SVFs by solving several systems of linear equations and update the meta-parameters through a two-level optimization structure to ensure its fast adaption and versatility. We validate our approach using real GPS data in Chengdu, China. Results show that our planned routes better match actual routes compared with state-of-the-art methods including the recursive logit model, Deep-IRL and Dij-IRL: the F1-Score increases by 4.17% with the introduction of the recursive logit model and further increases to 5.19% after meta-learning is employed. Moreover, we can reduce training time by over 95%.

Dynamic Recursive Logit Model for Vehicle Driving Route Choices and Path Inference with Incomplete Fixed Location Sensor Data

Dynamic Recursive Logit Model for Vehicle Driving Route Choices and Path Inference with Incomplete Fixed Location Sensor Data

Capturing positive network attributes during the estimation of recursive logit models: A prism-based approach

Although the recursive logit (RL) model has been recently popular and has led to many applications and extensions, an important numerical issue with respect to the computation of value functions remains unsolved. This issue is particularly significant for model estimation, during which the parameters are updated every iteration and may violate the feasibility condition of the value function. To solve this numerical issue of the value function in the model estimation, this study performs an extensive analysis of a prism-constrained RL (Prism-RL) model proposed by Oyama and Hato (2019), which has a path set constrained by the prism defined based upon a state-extended network representation. The numerical experiments have shown two important properties of the Prism-RL model for parameter estimation. First, the prism-based approach enables estimation regardless of the initial and true parameter values, even in cases where the original RL model cannot be estimated due to the numerical problem. We also successfully captured a positive effect of the presence of street green on pedestrian route choice in a real application. Second, the Prism-RL model achieved better fit and prediction performance than the RL model, by implicitly restricting paths with large detour or many loops. Defining the prism-based path set in a data-oriented manner, we demonstrated the possibility of the Prism-RL model describing more realistic route choice behavior. The capture of positive network attributes while retaining the diversity of path alternatives is important in many applications such as pedestrian route choice and sequential destination choice behavior, and thus the prism-based approach significantly extends the practical applicability of the RL model.

Undiscounted Recursive Path Choice Models: Convergence Properties and Algorithms

Traffic flow predictions are central to a wealth of problems in transportation. Path choice models can be used for this purpose, and in state-of-the-art models—so-called recursive path choice (RPC) models—the choice of a path is formulated as a sequential arc choice process using undiscounted Markov decision process (MDP) with an absorbing state. The MDP has a utility maximization objective with unknown parameters that are estimated based on data. The estimation and prediction using RPC models require repeatedly solving value functions that are solutions to the Bellman equation. Although there are several examples of successful applications of RPC models in the literature, the convergence of the value iteration method has not been studied. We aim to address this gap. For the two closed-form models in the literature—recursive logit (RL) and nested recursive logit (NRL)—we study the convergence properties of the value iteration method. In the case of the RL model, we show that the operator associated with the Bellman equation is a contraction under certain assumptions on the parameter values. On the contrary, the operator in the NRL case is not a contraction. Focusing on the latter, we study two algorithms designed to improve upon the basic value iteration method. Extensive numerical results based on two real data sets show that the least squares approach we propose outperforms two value iteration methods.

Capturing Positive Utilities During the Estimation of Recursive Logit Models: A Prism-Based Approach

Although the recursive logit (RL) model has been recently popular and has led to many applications and extensions, an important numerical issue with respect to the evaluation of value functions remains unsolved. This issue is particularly significant for model estimation, during which the parameters are updated every iteration and may violate the model feasible condition. To solve this numerical issue, this paper proposes a prism-constrained RL (Prism-RL) model that implicitly restricts the path set by the prism constraint defined based upon a state-extended network representation. Providing a set of numerical experiments, we show that the Prism-RL model succeeds in the stable estimation regardless of the initial and true parameter values and is able to capture positive utilities. In the real application to a pedestrian network, we found the positive effect of street green presence on pedestrians. Moreover, the Prism-RL model achieved higher goodness of fit than the RL model, implying that the Prism-RL model can also describe more realistic route choice behavior.

Impact of transport network disruption on travel demand: A case study of the July 2018 heavy rain disaster in Japan

This study analyzes the impacts of disruptions on transport network performance and travel demand in Hiroshima during the July 2018 heavy rain disaster. We first introduce a resilience concept to reflect the temporal dimension of the network recovery process explicitly. We then identify temporal changes in multimodal transport network performance using a simple logsum-based measure obtained from the recursive logit model. Because the recursive model is consistent with random utility theory, it also allows us to compute the additional cost of travel caused by the disaster. Finally, we explore the impacts of network performance loss on travel demand using mobile phone data. The main findings include (1) network disruption significantly reduced travel demand, and (2) the total monetary loss was around 5.7 billion JPY, which would be higher if traffic congestion were included. These results would help policymakers to facilitate the recovery process from transport network disruptions.

Spatial Transferability of Normalized Time-Discounted Recursive Logit Model in Excursion of Districts around Stations

Spatial Transferability of Normalized Time-Discounted Recursive Logit Model in Excursion of Districts around Stations

A recursive logit model with choice aversion and its application to transportation networks

We propose a recursive logit model which captures the notion of choice aversion by imposing a penalty term that accounts for the dimension of the choice set at each node of the transportation network. We make three contributions. First, we show that our model overcomes the correlation problem between routes, a common pitfall of traditional logit models, and that the choice aversion model can be seen as an alternative to these models. Second, we show how our model can generate violations of regularity in the path choice probabilities. In particular, we show that removing edges in the network may decrease the probability for existing paths. Finally, we show that under the presence of choice aversion, adding edges to the network can make users worse off. In other words, a type of Braess’s paradox can emerge outside of congestion and can be characterized in terms of a parameter that measures users’ degree of choice aversion. We validate these contributions by estimating this parameter over GPS traffic data captured on a real-world transportation network.

A Recursive Logit Model with Choice Aversion and Its Application to Route Choice Analysis

We propose a recursive logit model which captures the notion of choice aversion by imposing a penalty term that accounts for the dimension of the choice set at each node of the transportation network. We make three contributions. First, we show that our model overcomes the correlation problem between routes, a common pitfall of traditional logit models, and that the choice aversion model can be seen as an alternative to these models. Second, we show how our model can generate violations of regularity in the path choice probabilities. In particular, we show that removing edges in the network may decrease the probability for existing paths. Finally, we show that under the presence of choice aversion, adding edges to the network can make users worse off. In other words, a type of Braess's paradox can emerge outside of congestion and can be characterized in terms of a parameter that measures users' degree of choice aversion. We validate these contributions by estimating this parameter over GPS traffic data captured on a real-world transportation network.

Choice Aversion in Directed Networks

This paper studies the problem of optimal path selection in a directed network by decision makers that have an intrinsic distaste for evaluating too many options. We propose a recursive logit model that incorporates choice aversion along the lines introduced by Fudenberg and Strzalecki (2015). We derive optimal flow allocations both sequentially and from a path-choice perspective, which is robust to the presence of overlapping routes. We obtain a tight characterization of welfare in terms of both the network topology and the degree of choice aversion, where we derive comparative statics consistent with previous research.

Modelling intermodal travel in Switzerland: A recursive logit approach

We use data from the Swiss national household travel survey to 1. analyze the socioeconomic determinants for intermodal travel in Switzerland and 2. estimate a first large-scale multimodal recursive logit route choice model for urban trip making. We show that intermodal travel is mostly associated with ownership of transit subscriptions, which allow free at the point-of-use public transportation. We also present a framework using open-source data to generate the multimodal network for the recursive logit model estimation. The fact that the model only needs a multimodal network to model the transport supply makes it independent of path sampling algorithms for the choice-set construction and it thus provides an alternative to classic mode and route choice models, since it can estimate mode and route choice parameters with directly observed routes, removing the sampling bias. By eliminating the need to sample alternative paths for estimation, it also simplifies the estimation process, making it a viable choice as an integral solution for joint route and mode choice modelling.

Transferring Time-Series Discrete Choice to Link-Based Route Choice in Space: Estimating Vehicle Type Preference using Recursive Logit Model

This paper considers a sequential discrete choice problem in a time domain, formulated and solved as a route choice problem in a space domain. Starting from a dynamic specification of time-series discrete choices, we show how it is transferrable to link-based route choices that can be formulated by a finite path choice multinomial logit model. This study establishes that modeling sequential choices over time and in space are equivalent as long as the utility of the choice sequence is additive over the decision steps, the link-specific attributes are deterministic, and the decision process is Markovian. We employ the recursive logit model proposed in the context of route choice in a network, and apply it to estimate time-series vehicle type choice based on Maryland Vehicle Stated Preference Survey data. The model has been efficiently estimated by a dynamic programming approach; the values of estimated coefficients provide important patterns on vehicle type preferences. Compared with a naive model based on sequential multinomial logit choices which are independent over time and a dynamic discrete choice model which considers agent’s future expectations, the smaller root mean square error of recursive logit model indicates that it has a better performance in estimating sequential choices over time. We also compare the predictive powers and find that the proposed model outperforms the basic approach and the dynamic approach.

Capturing correlation with a mixed recursive logit model for activity-travel scheduling

Representing activity-travel scheduling decisions as path choices in a time–space network is an emerging approach in the literature. In this paper, we model choices of activity, location, timing and transport mode using such an approach and seek to estimate utility parameters of recursive logit models. Relaxing the independence from irrelevant alternatives (IIA) property of the logit model in this setting raises a number of challenges. First, overlap in the network may not fully characterize perceptual correlation between paths, due to their interpretation as activity schedules. Second, the large number of states that are needed to represent all possible locations, times and activity combinations imposes major computational challenges to estimate the model. We combine recent methodological developments to build on previous work by Blom Västberg et al. (2016) and allow to model complex and realistic correlation patterns in this type of network. We use sampled choices sets in order to estimate a mixed recursive logit model in reasonable time for large-scale, dense time-space networks. Importantly, the model retains the advantage of fast predictions without sampling choice sets. In addition to estimation results, we present an extensive empirical analysis which highlights the different substitution patterns when the IIA property is relaxed, and a cross-validation study which confirms improved out-of-sample fit.

A strategy-based recursive path choice model for public transit smart card data

A recursive logit model is proposed for path choice modeling with transit smart card data in higher-frequency bus and rail services. In such circumstances, it is commonly assumed that passengers may arrive randomly to a stop and may behave according to a “strategy”; such a strategy is associated with a so-called “attractive” set of routes: a passenger selects a set of routes departing from the stop, and will board the next vehicle to depart from among that set of routes. We extend the conventional notion of attractive sets by introducing a measure of “attractiveness” that allows for randomness in the choice of attractive routes.The proposed model uses a link-based (recursive, or sequential choice) formulation, rather than a path-based formulation, which has the advantage of including all path alternatives without the need for path set enumeration. The recursive formulation is also very suitable for smart card data because it allows model calibration with incomplete path choice observations.The link-based approach was originally advocated by Nguyen et al. (1988) in the strategy-based transit assignment context, but without investigating methods for model calibration or empirical analysis. Recently, Fosgerau et al. (2013) and Mai et al. (2015) have presented methods to empirically estimate traffic path choice models using a link-based formulation. Our framework builds off these previous works to formulate and estimate a strategy-based path choice model with smart card data. The proposed model is tested with a 6 months extract from the smart card transactions in Brisbane, Australia, for two popular origin-destination pairs with diverse path alternatives.

Operational route choice methodologies for practical applications

This paper focuses on the application of tractable route choice models and presents a set of methods for deriving relevant disaggregate and aggregate route choice indicators, namely link and route flows. Tractability is achieved at the disaggregate level by the recursive logit model and at the aggregate level by the mental representation item ($$\mathrm {MRI}$$) approach. These two approaches are analyzed here, and extensions of the $${\mathrm {MRI}}$$ approach are presented. The analysis elaborates on the features of each model and allows to draw insights into the use of a specific model, depending on the needs of the application and the data availability. The performance of the two models is tested on real data. The results demonstrate the validity of the $${\mathrm {MRI}}$$ model that is intended for aggregate analysis.

Route Choice Analysis in the Tokyo Metropolitan Area Using a Link-based Recursive Logit Model Featuring Link Awareness

Abstract Identification of an appropriate route choice model to understand travel behavior remains challenging. To this end, Fosgerau et al. (2013) have recently developed a link-based route choice model termed the “recursive logit” (RL) model. A decision-maker is assumed to choose the next link recursively that maximizes the sum of instantaneous utility and expected downstream utility at each node. However, in practical application, some computational issues remain, including large (and often ill-defined) matrix inversions. Here, we develop an alternative RL model that considers the probability of awareness of the next link that improves the stability of model estimations. The model was estimated using vehicle trajectory data from the ETC (Electronic Toll Collection) 2.0 dataset of the Tokyo Metropolitan area, and the results were compared to those of a conventional RL model in terms of predictive accuracy and computational efficiency.

A discounted recursive logit model for dynamic gridlock network analysis

Emerging sensing technologies such as probe vehicles equipped with Global Positioning System (GPS) devices on board provide us real-time vehicle trajectories. They are helpful for the understanding of the cases that are significant but difficult to observe because of the infrequency, such as gridlock networks. On the premise of this type of emerging technology, this paper propose a sequential route choice model that describes route choice behavior, both in ordinary networks, where drivers acquire spatial knowledge of networks through their experiences, and in extraordinary networks, which are situations that drivers rarely experience, and applicable to real-time traffic simulations. In extraordinary networks, drivers do not have any experience or appropriate information. In such a context, drivers have little spatial knowledge of networks and choose routes based on dynamic decision making, which is sequential and somewhat forward-looking. In order to model these decision-making dynamics, we propose a discounted recursive logit model, which is a sequential route choice model with the discount factor of expected future utility. Through illustrative examples, we show that the discount factor reflects drivers’ decision-making dynamics, and myopic decisions can confound the network congestion level. We also estimate the parameters of the proposed model using a probe taxis’ trajectory data collected on March 4, 2011 and on March 11, 2011, when the Great East Japan Earthquake occurred in the Tokyo Metropolitan area. The results show that the discount factor has a lower value in gridlock networks than in ordinary networks.

一般化 RL モデルを用いた災害時の経路選択行動分析

一般化 RL モデルを用いた災害時の経路選択行動分析