Corpus ID: 237635198

Combining Discrete Choice Models and Neural Networks through Embeddings: Formulation, Interpretability and Performance

  title={Combining Discrete Choice Models and Neural Networks through Embeddings: Formulation, Interpretability and Performance},
  author={Ioanna Arkoudi and Carlos Lima Azevedo and Francisco C. Pereira},
This study proposes a novel approach that combines theory and data-driven choicemodels using Artificial Neural Networks (ANNs). In particular, we use continuous vector representations, called embeddings, for encoding categorical or discrete explanatory variables with a special focus on interpretability and model transparency. Although embedding representations within the logit framework have been conceptualized by Pereira in [36], their dimensions do not have an absolute definitive meaning… Expand

Figures and Tables from this paper


A Neural-embedded Choice Model: TasteNet-MNL Modeling Taste Heterogeneity with Flexibility and Interpretability
TasteNet-MNL can recover the underlying non-linear utility function, and provide predictions and interpretations as accurate as the true model; while examples of logit or random coefficient logit models with misspecified utility functions result in large parameter bias and low predictability. Expand
Rethinking travel behavior modeling representations through embeddings
The concept of travel behavior embeddings is introduced, a method for re-representing discrete variables that are typically used in travel demand modeling, such as mode, trip purpose, education level, family type or occupation, and a new Python package, called PyTre, that others can straightforwardly use to replicate the results or improve their own models. Expand
Neural networks and the multinomial logit for brand choice modelling: a hybrid approach
It is shown that a Feedforward Neural Network with Softmax output units and shared weights can be viewed as a generalization of the Multinomial Logit model, which is used as a diagnostic and specification tool for the Logit, which will provide interpretable coefficients and significance statistics. Expand
An empirical comparison of the validity of a neural net based multinomial logit choice model to alternative model specifications
This work analyzes purchase data of the six largest brands in terms of market share for two product groups and finds inversely S-shaped, saturation and interaction effects on utility, and shows that the increase in complexity caused by the neural choice model is justified by higher validity. Expand
A flexible brand choice model based on neural net methodology A comparison to the linear utility multinomial logit model and its latent class extension
This feedforward multilayer perceptron is able to approximate any continuous multivariate function and its derivatives with the desired level of precision and leads to better out-of-sample results than homogeneous and heterogeneous versions of linear utility MNL models. Expand
Hybrid Choice Models: Progress and Challenges
We discuss the development of predictive choice models that go beyond the random utility model in its narrowest formulation. Such approaches incorporate several elements of cognitive process thatExpand
Asymmetric, closed-form, finite-parameter models of multinomial choice
In transportation, the number of observations associated with one discrete outcome is often greatly different from the number of observations associated with another discrete outcome. This situationExpand
Why did you predict that? Towards explainable artificial neural networks for travel demand analysis
It is shown that by reconceptualising the LRP methodology towards the choice modelling and travel demand analysis contexts, it can be put to effective use in application domains well beyond the field of computer vision, for which it was originally developed. Expand
High Dimensional Nonparametric Discrete Choice Model
The functional form of a model can be a constraint in the correct prediction of discrete choices. The fl exibility of a nonparametric model can increase the likelihood of correct prediction. TheExpand
node2vec: Scalable Feature Learning for Networks
In node2vec, an algorithmic framework for learning continuous feature representations for nodes in networks, a flexible notion of a node's network neighborhood is defined and a biased random walk procedure is designed, which efficiently explores diverse neighborhoods. Expand