Code for the article "Does causal reasoning help preventing churn?", co-authored by Théo Verhelst, Olivier Caelen, Jean-Christophe Dewitte, and Gianluca Bontempi.
- Implementation of the true and estimated uplift curve. The true uplift curve can be used in simulation settings, whereas its estimated counterpart can be used in empirical settings.
- Implementation of the causal precision curve (see article).
- Hierarchical bayesian generative model of customer churn, which can be parametrized to simulate any distribution of customer types
- Example of use.
- R (>= 3.4.1)
ggplot2
for the examples