Answer 1 and Question 2¶
Identify poor households eligible for welfare program based on the house characteristics (#rooms, sqm, neighborhood) and observable durable goods (tv, fridge, toilet).
Prediction
Why? You need a training data where you know exactly the poverty status of each family plus the house characteristics and other observables. Then, you train the model to predict the poverty status based on the predictors. You are not interested in whether the prediction model uses the number of tv-sets or toilets to predict poverty. You are interested how accurately these predictors when working together can predict poverty.
Question 2¶
Find which advertisement campaign is the most effective for boosting product sales.
Is it a prediction task or an inference task?
(See the answer in the next section)