Inferring and Modeling Migration Flows Using Mobile Phone CDR Data

Abstract

Understanding the causes and impacts of migration, as well as implementing policies aimed at providing certain services, requires estimating migration flows and forecasting future patterns. Over time, less study has been done on modeling migration flows than has been done on modeling other types of flows, such as commutes. One of the biggest hurdles to empirical analysis and theoretical developments in the modeling of migration flows has been a lack of data. Because a migration trip is far less frequent than a commute, it necessitates a longitudinal set of data for study. The data from a large mobile phone network is used in this chapter to infer migration trips and their distribution. Intra/inter-district migration flows, migration distance distribution, and origin-destination (O-D) movements are among the interesting properties of the inferred migration trips. The log-linear model, classic gravity model, and recently developed radiation model are investigated for migration trip distribution modeling, with distinct approaches applied in setting parameters for each model. As a result, among the different models, gravity and log-linear models with a direct distance (displacement) as a trip cost and district centroids as reference points perform the best. Among the radiation models, a model that considers district population is the best performing model, but not as good as the gravity and log-linear models. This chapter reflects the idea and thinking process of our original work by Phithakkitnukoon et al. (IEEE Access. 2022;10:23248–58; IEEE international conference on privacy, security, risk and trust and IEEE international conference on social computing (PASSAT/SocialCom 2011); 2011. p. 515–20), and Hankaew et al. (IEEE Access. 2019;7(1):164746–58).