The Residence History Inference Problem
The use of online user traces for studies of human mobility has received significant attention in recent years. This growing body of work, and the more general importance of human migration patterns to government and industry, motivates the need for a formalized approach to the computational modeling of human mobility - in particular how and when individuals change their place of residence - from online traces. Prior work on this topic has skirted the underlying computational modeling of residence inference, focusing on migration patterns themselves. As a result, to our knowledge, all prior work has employed heuristics to compute something like residence histories. Here, we formalize the residence assignment problem, which seeks, under constraints associated with the minimum length-of-stay at a residence, the most parsimonious sequence of residence periods and places that explains the movement history of an individual. Here we provide an exact solution for this problem and establish its algorithmic complexity. Because the calculation of optimal residence histories (under the assumptions of the model) is tractable, we believe that this method will be a valuable tool for future work on this topic.
READ FULL TEXT