Most machine learning problems have some element of location involved, whether it is the zip code of a customer, a place name description, address of a business, or latitude and longitude of a shipment.
However, working with geospatial data correctly is difficult. Typically, practitioners ignore geospatial information, or more advanced data scientists put in the effort to handle location data manually–for example, enriching with other data sources, generating neighbor features, and creating spatial clusters. Is this effort worth it?
