A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Exploring graph neural networks for informal settlements classification
Informal settlements are a global phenomenon, resulting from rapid urbanization and limited infrastructure. These areas are characterized by irregular land tenure and lack of infrastructure, representing significant challenges to urban planning. Innovative approaches such as deep learning techniques have been employed over the past several years to analyze city dynamics and understand the complexity of informal settlements to address the challenges of achieving sustainable urban planning. However, many cities, lack surveys collecting data on legal tenancy rights. This paper explores a potential solution to the absence of this type of data by developing a label propagation model to classify neighborhoods based on geographic and satellite information, which is then used as input to a second model that estimates the percentage of informal areas employing demographic and health data. The study applied a Graph Convolutional Network (GCN) to classify each neighborhood as either “legalized” or “informal”, and a second GCN to predict informal areas at the Urban Zonnal Planning Unit (UPZ) level, both using information from Bogota, Colombia. The initial model achieved an AUC of 88.89% and the second model successfully predicted 80% of UPZ values with an error below 10% for a mean absolute error (MAE) of 7.06%. However, this last model struggled to predict UPZs with high informality levels. These results provide a starting point for applying GCNs in urban planning contexts related to informal settlements and lay the foundation for scaling these models to other cities to improve urban planning strategies globally. ; Maestría
Exploring graph neural networks for informal settlements classification
Informal settlements are a global phenomenon, resulting from rapid urbanization and limited infrastructure. These areas are characterized by irregular land tenure and lack of infrastructure, representing significant challenges to urban planning. Innovative approaches such as deep learning techniques have been employed over the past several years to analyze city dynamics and understand the complexity of informal settlements to address the challenges of achieving sustainable urban planning. However, many cities, lack surveys collecting data on legal tenancy rights. This paper explores a potential solution to the absence of this type of data by developing a label propagation model to classify neighborhoods based on geographic and satellite information, which is then used as input to a second model that estimates the percentage of informal areas employing demographic and health data. The study applied a Graph Convolutional Network (GCN) to classify each neighborhood as either “legalized” or “informal”, and a second GCN to predict informal areas at the Urban Zonnal Planning Unit (UPZ) level, both using information from Bogota, Colombia. The initial model achieved an AUC of 88.89% and the second model successfully predicted 80% of UPZ values with an error below 10% for a mean absolute error (MAE) of 7.06%. However, this last model struggled to predict UPZs with high informality levels. These results provide a starting point for applying GCNs in urban planning contexts related to informal settlements and lay the foundation for scaling these models to other cities to improve urban planning strategies globally. ; Maestría
Exploring graph neural networks for informal settlements classification
2024-12-05
Theses
Electronic Resource
English
Transforming informal settlements to sustainable settlements
British Library Online Contents | 2002
Informal Settlements and Human Health
| Springer Verlag | 2018
Property rights in informal settlements
| Elsevier | 2021
Circular Economy in Informal Settlements
| Taylor & Francis Verlag | 2023
Generative Modelling of Informal Settlements
| British Library Conference Proceedings | 1994