Measuring Resilient Communities: An Analytical And Predictive Tool

Silvio Carta University of Hertfordshire
Tommaso Turchi University of Hertfordshire
Luigi Pintacuda University of Hertfordshire

This work presents the initial results of an analytical tool designed to quantitatively assess the level of resilience of urban areas. We use Deep Neural Networks to extract features of resilience from a trained model that classifies urban areas using a pre-assigned value range of resilience. The model returns the resilience value for any urban area, indicating the distance between the centre of the selected area and relevant typologies, including green areas, buildings, natural elements and infrastructures. Our tool also indicates the urban morphological characteristics that have a larger impact on the resilience score. In this way we can learn why a neighbourhood is successful (or not) and how to improve its level of resilience. The model employs Convolutional Neural Networks (CNNs) with Keras on Tensorflow for the computation. The outputs are loaded onto a Node.js environment and bootstrapped with React.js to generate the online demo.

Keywords: Sustainable Cities And Communities, Resilient Communities, Cnn, Urban Morphology, Sdg 11, Sdg 13

View Full Paper