Cabanas-Tirapu, O., Danús, L., Moro, E. et al. Human mobility is well described by closed-form gravity-like models learned automatically from data. Nat Commun

Objective:

• Compare the performance at predicting mobility flows of simple gravity models, complex machine learning and deep-learning methods, and closed-form, interpretable models obtained through Bayesian symbolic regression

Case:

• Six states, US

Methodology:

• Flow prediction:
  • Bayesian machine scientist
  • Gravity model
  • Radiation model
  • Random forest
  • Deep gravity

Data Source

• Mobile phone

Findings:

• Automated equation discovery approaches lead to parsimonious closed-form models that combine the most desirable aspects

Coding Reference:

• Python