Political Violence in Athens, Greece (2008-2024): A Machine Learning Approach for Predictive Modelling of Spatial Risk Patterns
(2025) In Master Thesis in Geographical Information Science GISM01 20252Dept of Physical Geography and Ecosystem Science
- Abstract
- Political violence in Athens exhibits distinctive spatial and temporal regularities shaped by the city’s symbolic infrastructure, mobility patterns, and socio-political dynamics. This thesis integrates environmental criminology with spatially explicit machine-learning techniques to examine risk patterns of politically motivated violence between 2008 and 2024. Using a multi-layered dataset of 610 attack events and 610 matched non-events, the study develops a three-tier modelling framework grounded in Random Forests (RF), and enriched with coefficients derived from Multiscale Geographically Weighted Regression (MGWR) and Geographically and Temporally Weighted Regression (GTWR).
The baseline RF model captures non-linear relationships between... (More) - Political violence in Athens exhibits distinctive spatial and temporal regularities shaped by the city’s symbolic infrastructure, mobility patterns, and socio-political dynamics. This thesis integrates environmental criminology with spatially explicit machine-learning techniques to examine risk patterns of politically motivated violence between 2008 and 2024. Using a multi-layered dataset of 610 attack events and 610 matched non-events, the study develops a three-tier modelling framework grounded in Random Forests (RF), and enriched with coefficients derived from Multiscale Geographically Weighted Regression (MGWR) and Geographically and Temporally Weighted Regression (GTWR).
The baseline RF model captures non-linear relationships between environmental and infrastructural predictors and political violence risk, while the MGWR-informed variant adds local spatial heterogeneity, improving interpretability and classification performance. GTWR-based enrichments further reveal that the spatial logic of political violence changed significantly across periods: the 2008-2016 crisis era shows more structured and centralised targeting patterns, whereas post-2017 incidents become increasingly diffuse. Across models, predictors related to symbolic targets, accessibility, and guardianship consistently shape risk.
The findings demonstrate that combining environmental criminology with spatial machine learning enhances both explanatory depth and predictive accuracy. The study highlights the value of dynamic modelling approaches for identifying evolving opportunity structures and informing context-sensitive urban security planning. Ultimately, this research advances the theoretical integration of criminological concepts with GeoAI, illustrating how spatial and temporal mechanisms jointly underpin the geography of political violence in urban environments. (Less) - Popular Abstract
- Political violence has been a recurring feature of urban life in parts of Athens for many years, yet the question of where such incidents are more likely to occur has received far less attention than the question of why they happen. This thesis explores how features of the city itself – its symbolic locations, public infrastructure, and patterns of accessibility – help shape the geography of politically motivated violence.
By combining core concepts from environmental criminology with advanced spatial analysis and machine-learning techniques, the research examined incidents that took place between 2008 and 2024 and compared them with a wide range of environmental and urban indicators. To this end, a set of predictive models was developed... (More) - Political violence has been a recurring feature of urban life in parts of Athens for many years, yet the question of where such incidents are more likely to occur has received far less attention than the question of why they happen. This thesis explores how features of the city itself – its symbolic locations, public infrastructure, and patterns of accessibility – help shape the geography of politically motivated violence.
By combining core concepts from environmental criminology with advanced spatial analysis and machine-learning techniques, the research examined incidents that took place between 2008 and 2024 and compared them with a wide range of environmental and urban indicators. To this end, a set of predictive models was developed to compare how political violence is distributed across space and how these patterns change over time. To ensure the analysis remained transparent and interpretable, the models incorporated spatial context and temporal change while being evaluated with multiple diagnostic tools.
The findings show that patterns of political violence in Athens are far from random: they tend to cluster in areas where symbolic buildings, transportation networks, and dense urban activity come together. These patterns also evolve over time, with earlier years displaying more concentrated spatial behaviours and more recent years showing a tendency toward dispersal across the city.
Finally, the study underscores how characteristics of the urban environment can influence where political violence is more likely to emerge. Emphasising the value of spatial data and responsible analytical methods, these results offer a data-informed perspective that can support urban safety, planning, and resilience, helping cities understand how risk patterns develop and transform. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9216047
- author
- Andriotis, Symeon LU
- supervisor
- organization
- course
- GISM01 20252
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Geography, GIS, Geographical Information Systems, Environmental Criminology, Political Violence, GeoAI, Machine Learning, Random Forests, Multiscale Geographically Weighted Regression, Geographically and Temporally Weighted Regression, Predictive Policing, Spatial Risk Modelling, Athens, Greece
- publication/series
- Master Thesis in Geographical Information Science
- report number
- 202
- language
- English
- id
- 9216047
- date added to LUP
- 2025-12-08 16:21:44
- date last changed
- 2025-12-08 16:21:44
@misc{9216047,
abstract = {{Political violence in Athens exhibits distinctive spatial and temporal regularities shaped by the city’s symbolic infrastructure, mobility patterns, and socio-political dynamics. This thesis integrates environmental criminology with spatially explicit machine-learning techniques to examine risk patterns of politically motivated violence between 2008 and 2024. Using a multi-layered dataset of 610 attack events and 610 matched non-events, the study develops a three-tier modelling framework grounded in Random Forests (RF), and enriched with coefficients derived from Multiscale Geographically Weighted Regression (MGWR) and Geographically and Temporally Weighted Regression (GTWR).
The baseline RF model captures non-linear relationships between environmental and infrastructural predictors and political violence risk, while the MGWR-informed variant adds local spatial heterogeneity, improving interpretability and classification performance. GTWR-based enrichments further reveal that the spatial logic of political violence changed significantly across periods: the 2008-2016 crisis era shows more structured and centralised targeting patterns, whereas post-2017 incidents become increasingly diffuse. Across models, predictors related to symbolic targets, accessibility, and guardianship consistently shape risk.
The findings demonstrate that combining environmental criminology with spatial machine learning enhances both explanatory depth and predictive accuracy. The study highlights the value of dynamic modelling approaches for identifying evolving opportunity structures and informing context-sensitive urban security planning. Ultimately, this research advances the theoretical integration of criminological concepts with GeoAI, illustrating how spatial and temporal mechanisms jointly underpin the geography of political violence in urban environments.}},
author = {{Andriotis, Symeon}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master Thesis in Geographical Information Science}},
title = {{Political Violence in Athens, Greece (2008-2024): A Machine Learning Approach for Predictive Modelling of Spatial Risk Patterns}},
year = {{2025}},
}