Quarantines and lockdowns due to COVID-19 have had multiple social effects in terms of family isolation, economic uncertainty and general stress that led to a potential increase in domestic violence. In Bogotá, Colombia, official reports show an increase of 20.3 percent (2,702 cases) in the number of domestic violence reports from January to May 2020, relative to the same period in 2019. Information on imposed mobility restrictions and quantified mobility changes is used to develop a spatial-temporal domestic violence occurrence prediction model accompanied by machine learning interpretation techniques to understand the association of mobility patterns with the spatial-temporal behaviour of domestic violence incidents in Bogotá. Additionally, different OLS-based regression models are estimated to better understand the relationship between the duration of the quarantine lockdown and domestic violence, which was initially found to be significant from the predictive spatial-temporal approximation.

This research is presented as a joint work by researchers from Quantil | Applied Mathematics and the Office of Information Analysis and Strategic Studies of the District Secretariat of Security, Coexistence and Justice for Bogotá. These institutions have previously worked together as part of a research project funded by Colciencias on crime prediction models for robberies and homicides in Bogotá.