Causal modeling is an umbrella time period for a variety of strategies that enable us to mannequin the consequences of our actions on the world.
Causal fashions differ from conventional machine studying fashions in quite a lot of methods.
A very powerful distinction between them stems from the truth that the knowledge contained in observational information used to coach conventional machine studying equipment is — basically — inadequate to persistently mannequin the consequences of our actions.
The end result?
Utilizing conventional machine studying strategies to mannequin the outcomes of our actions results in biased choices more often than not.
An excellent instance right here is utilizing a regression mannequin skilled on historic information to find out your advertising and marketing combine.
One other one?
Utilizing XGBoost skilled on historic observations to foretell the chance of churn and sending a marketing campaign if the anticipated chance of churn is larger than some threshold.