ABSTRACT
In Part 1, the predictions omit the ‘spatial process’ of home prices – namely that prices cluster near one another. In Part 2, readers learn how to engineer new features to account for this spatial process and then open the black box of their machine learning model to understand whether predictions generalize across different urban contexts, like low vs. high income neighborhoods and white vs. minority neighborhoods. The chapter finishes with a first glimpse into algorithmic fairness, a central theme of the book in the coming chapters.
