Spatio-Temporal Modeling and Crisis Sensibility Analysis Using Social Media Data

Details

Supervisors

Hafner, Christian ; Dujardin, Sébastien

Faculty

Faculté des sciences

Degree label

Master [120] en science des données, orientation statistique, à finalité spécialisée

Abstract

This thesis explores spatio-temporal modeling techniques to analyze social media user metrics, using Meta's data as a proxy for population distribution in the Philippines. The study examines various approaches, including econometric models, kriging, and Bayesian modeling for Gaussian processes, to address spatial and temporal auto-correlations in the data. Kernel-based methods are utilized both in spatio-temporal modeling, particularly through Gaussian processes, and in the development of a dynamic Crisis Sensitivity Index (CSI) during typhoon events. Overall, the thesis demonstrates the effectiveness of these advanced techniques in handling complex spatio-temporal datasets for population analysis and crisis impact assessment.