Modelling Passenger Behaviour on the Underground

Explore a 37-minute lecture on modeling passenger behavior in the London Underground, presented by the Alan Turing Institute. Delve into two main research areas: an end-to-end model for inferring changes in passenger behavior during unplanned disruptions, incorporating user-level data and heterogeneous choice inferred from smart card outputs, and an abstract view of causal models for complex systems with shock records and observational data. Learn about network tomography problems, transportation networks challenges, and experimental results using TfL datasets for travel time prediction and station load estimation. Gain insights into gradient backpropagation, intractability, and approximation techniques used in these models.