This repository contains the code for various uplift modelling algorithms, including:
- S-learner
- T-learner
- Simplified X-learner
- X-learner
Uplift modelling is a causal inference technique that predicts the incremental impact of a treatment on an individual's behaviour, which allows businesses to target only those customers who are likely to respond positively to a treatment. This can be used to improve the effectiveness of marketing campaigns, reduce costs, and increase revenue.
By definition, there will be four possible outcomes based on a response to a treatment:
- Persuadable
- Sure Thing
- Lost Cause
- Do Not Disturb
In practice, the actual definition of these outcomes may vary depending on the hypothesis. When predicting uplift score and selecting a segment by the highest score, we are trying to find the only one type of customer: persuadable in revenue uplift or do not disturb in churn uplift.