We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Dr. Juan Orduz: Introduction to Uplift Modeling
"Join Dr. Juan Orduz for an introduction to uplift modeling, exploring treatment effect estimation, common models like SL, TL, and XL, and essential considerations for accounting for user variations and ensuring the treatment effect is isolated."
- Uplift modeling aims to estimate the treatment effect, calculating how much a treatment (in this case, an email) moves users from one state to another.
- Common uplift models include S learner (SL), T learner (TL), and X learner (XL), all estimating the treatment effect.
- A perfect model assigns higher scores to all treated individuals with positive outcomes and vice versa.
- The X axis plots the covariates, whereas the Y axis plots the outcome.
- Different users have different responses to a treatment, making it essential to account for variations.
- Randomized designs help ensure the treatment assignment is not influenced by the outcome.
- Modeling must account for the unconfounded assumption, ensuring the treatment effect is isolated from other factors.
- The first model, S learner (SL), calculates the imputed treatment effect and then uses it for targeting.
- The second model, T learner (TL), calculates the treatment effect individually.
- The third model, X learner (XL), combines the strengths of S learner and T learner.
- Other issues to consider include repeated testing and bias against specific groups or churn.