UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Predicting Non-Residential Building Fire Risk Using Geospatial Information and Convolutional Neural Networks

Anderson-Bell, J; Schillaci, C; Lipani, A; (2021) Predicting Non-Residential Building Fire Risk Using Geospatial Information and Convolutional Neural Networks. Remote Sensing Applications: Society and Environment , 21 , Article 100470. 10.1016/j.rsase.2021.100470. Green open access

[thumbnail of R1_pre-print_Pred Non-Residential Building Fire.pdf]
Preview
Text
R1_pre-print_Pred Non-Residential Building Fire.pdf - Accepted Version

Download (2MB) | Preview

Abstract

Building fire risk prediction is crucial for allocation of building inspection resources and prevention of fire incidents. Existing research of building fire prediction makes use of data relating to local demography, crime, building use and physical building characteristics, yet few studies have analysed the relative importance of predictive features. Furthermore, image features relating to buildings, such as aerial imagery and digital surface models (DSM), have not been explored. This research presents a multi-modal hybrid neural network for the prediction of fire risk at the building level using the London Fire Brigade dataset. The inclusion of traditional and novel image features is assessed using Shapley values and an ablation study. The ablation study found that while building use is the most effective contributor of classification performance, demographic features, apart from social class, are detrimental. Moreover, while the DSM did not lead to any notable improvement in classification performance, the inclusion of the aerial imagery feature lead to a 4% increase in median validation ROC AUC. The final model presented achieved an ROC AUC of 0.8195 on the test set.

Type: Article
Title: Predicting Non-Residential Building Fire Risk Using Geospatial Information and Convolutional Neural Networks
Open access status: An open access version is available from UCL Discovery
DOI: 10.1016/j.rsase.2021.100470
Publisher version: https://doi.org/10.1016/j.rsase.2021.100470
Language: English
Additional information: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Civil, Environ and Geomatic Eng
URI: https://discovery.ucl.ac.uk/id/eprint/10119262
Downloads since deposit
286Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item