Inference (causal) models

Machine Learning for Business

Karolis Urbonas

Head of Machine Learning & Science, Amazon

What is causality?

• Identify causal relationship of how much certain actions affect an outcome of interest
• Answers the "why?" questions
• Optimizes for model interpretability vs. performance
• Models try to detect patterns in observed data (observational) and draw causal conclusions

Experiments vs. observations

• Experiments are designed and causal conclusions are guaranteed e.g. in A/B tests
• When experiments are impossible (unethical, too expensive, both) - the models are used (also called observational studies) to calculate effect of certain inputs on desired outcomes
• Experiments are always preferred over observational studies whenever possible

Best practices

1. Do experiments wherever you can
2. If running experiments all the time is too expensive, run them periodically (quarterly, annually) and use it as benchmark
3. If there are no way to run any experiments, build a causal model. This will require an advanced methodology

Inference model example

Inference - training

Inference - learning

Inference - regression coefficients

Inference - interpretation

Let's practice!

Machine Learning for Business