Neural Network for Discrete Choice: Modeling, Statistics, and Computation

Discrete choice models are a core tool in operations, marketing, and economics for understanding consumer behaviour, supporting demand forecasting and revenue optimisation. However, learning these models remains challenging due to the complexity of human decision-making. Motivated by recent advances in AI, we leverage neural networks (NNs) to model rich behavioural nuances, particularly consumer taste heterogeneity and bounded rationality. Our frameworks harmonise the flexibility and scalability of NNs with strong theoretical soundness, such as generalisation guarantees and global algorithmic convergence. Empirical studies substantiate our theoretical results, and our algorithmic framework also sheds light on other operations problems, such as assortment optimisation.