Netflix/sherlock: Causal Machine Learning for Segment Discovery and Analysis

Population segment discovery and analysis using state-of-the-art, doubly robust estimation techniques rooted in causal inference and machine learning. Using data from randomized A/B experiments or quasi-experiments (observational studies), a set of candidate segmentation variables are used to discover segments of the study population based on estimated treatment response heterogeneity, characterized by the conditional average treatment effect. Implemented procedures for estimation of the conditional average treatment effect incorporate ensemble machine learning (or the user's choice of regression algorithms) via the Super Learner ensemble modeling procedure in 'sl3', available for download from GitHub using 'remotes::install_github("tlverse/sl3")'.

install.packages("remotes")
remotes::install_github("Netflix/sherlock")