Causal ML
A/B Testing and Causal Inference are stuck in the dark ages. Enter Causal ML - a new suite of uplift modeling and causal inference methods. Let's dive in.
1. Causal machine learning (ML) refers to a branch of machine learning that focuses on understanding and modeling the causal relationships in data, rather than just finding correlations.
2. Causal ML is particularly useful in decision-making scenarios where understanding the cause of an event is more important than just predicting its occurrence. Typical use case in business: Did spending money on online advertising increase sales?
3. This often involves Causal Inference, Counterfactual Reasoning, Causal Discovery, and Intervention Effect Estimation. Let's break these terms down.
4. Causal Inference: Determining whether a relationship between two variables is causal or merely correlational. This often involves statistical methods that attempt to control for confounding variables.
5. Counterfactual Reasoning: Understanding what would happen to one variable if you changed another, often framed as “what if” questions. For instance, “What would happen to hotel cancelations outcomes if lead time (time between purchase and hotel stay) was reduced?”
6. Causal Discovery: Identifying causal relationships from data without prior knowledge about the potential connections between variables.
7. Intervention and Treatment Effects: Estimating the impact of interventions (like a new advertisement or email campaign) on an outcome.
CausalML Package: https://causalml.readthedocs.io/en/latest/about.html
===
Want to learn Causal ML in #Python?
I have a free workshop where I will share how to use the CausalML library, which is used by big companies like Uber, Google, and more for their online experiments.