Machine Learning Tools for the Analysis of Migration Trajectories

Barcelona, Spain & Hybrid, 13 & 15 June 2023

The LIFELONGMOVE project and the IUSSP Scientific Panel on Lifetime Migration organized two training sessions on “Machine Learning Tools for the Analysis of Migration Trajectories” in collaboration with the Centre for Demographic Studies at the Autonomous University of Barcelona. The training was offered in a hybrid mode and was attended by around 80 scholars.

The first 2-hour session on June 13, 2023, was led by IUSSP Panel co-chair Prof. Sergi Vidal, who offered an introduction to sequence analysis. The aim was to provide attendees with basic knowledge on how to analyse migration trajectories over the life course and their possible association with individual and societal well-being. The session covered details on how to handle sequence data in the R environment, using data from a sample of 802 residents in the Greater Santiago Metropolitan Area (Chile) as an example.

The second session on June 15, 2023, was led by Prof. Danilo Bolano, Research Fellow and Lecturer in Social Demography at Bocconi University Carlo F. Dondena Centre for Research on Social Dynamics. He discussed machine learning approaches (Boruta feature selection algorithm, and Lasso stability selection) to cluster life trajectories in R based on the same dataset on Santiago’s resident biographies. At the end of his 2-hour session, Prof. Bolano stressed how feature selection algorithms might represent a significant improvement to classical model-based regression approaches in social sciences, given their ability to deal with highly collinear sets of variables.

Course materials (handouts, datasets, and RStudio codes) were made available to participants prior to the start of the training session so as to enable them to follow the empirical testing step by step. Materials can be accessed upon request at lifelongmove@ced.uab.es.