Using virtual inheritance.

Assume the renting out is driver by a latent real variable \(Z\)

Also assume that Z is a function of the tax paid and No of occupants.

You interpret rented out bool as cummulative probability of this hidden latent renting indicator.

So your model is basically

Split the data into train and test. Fit model using train set. Get model uncertainty estimates from test data.

Assuming the model shows low uncertainty -- identify those bourgeois piggy banks.