A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Predicting Road Network Vulnerability to Fluvial Flooding Using Machine Learning Classifiers: Case Study of Houston during Hurricane Harvey
The objective of this study is to identify vulnerable sections in the transportation network with the help of machine learning classifiers. Many network-theory based frameworks have been proposed to assess the vulnerability of transportation networks using network centrality based measures; however, those measures can not be directly translated into the actual vulnerability of transportation networks as many studies seem to proclaim. This is because vulnerability is not only about the failure consequence but also failure probability, and there are clear heterogeneities in disaster-exposure levels of the individual nodes due to the spatially embedded nature of transportation networks. It is possible to study and characterize this heterogeneity with the help of classification tools in machine learning. First, the road network at a super neighborhood level is modeled as a primal graph. Then, a new measure for flood exposure of the nodes in a road network was proposed and treated as the dependent variable. Two independent variables, namely elevation and the shortest distance from flood control infrastructure, were identified for each individual node. A classification algorithm was trained and tested in order to predict the flood exposure of individual nodes in the road network. In the end, connectivity of the road network was estimated after removing nodes (which are predicted using the best performing classification algorithm) that are particularly vulnerable to fluvial flooding. The results indicated that the K-means clustering algorithm had the highest prediction accuracy. The proposed methodology was then applied to assess the vulnerability of other super neighborhoods in Houston during Hurricane Harvey. The proposed framework expands the scope of traditional vulnerability assessment analysis for the road networks by effectively making use of machine learning tools, as well as publicly available data. Results from this study could be used to inform resilience enhancement decisions.
Predicting Road Network Vulnerability to Fluvial Flooding Using Machine Learning Classifiers: Case Study of Houston during Hurricane Harvey
The objective of this study is to identify vulnerable sections in the transportation network with the help of machine learning classifiers. Many network-theory based frameworks have been proposed to assess the vulnerability of transportation networks using network centrality based measures; however, those measures can not be directly translated into the actual vulnerability of transportation networks as many studies seem to proclaim. This is because vulnerability is not only about the failure consequence but also failure probability, and there are clear heterogeneities in disaster-exposure levels of the individual nodes due to the spatially embedded nature of transportation networks. It is possible to study and characterize this heterogeneity with the help of classification tools in machine learning. First, the road network at a super neighborhood level is modeled as a primal graph. Then, a new measure for flood exposure of the nodes in a road network was proposed and treated as the dependent variable. Two independent variables, namely elevation and the shortest distance from flood control infrastructure, were identified for each individual node. A classification algorithm was trained and tested in order to predict the flood exposure of individual nodes in the road network. In the end, connectivity of the road network was estimated after removing nodes (which are predicted using the best performing classification algorithm) that are particularly vulnerable to fluvial flooding. The results indicated that the K-means clustering algorithm had the highest prediction accuracy. The proposed methodology was then applied to assess the vulnerability of other super neighborhoods in Houston during Hurricane Harvey. The proposed framework expands the scope of traditional vulnerability assessment analysis for the road networks by effectively making use of machine learning tools, as well as publicly available data. Results from this study could be used to inform resilience enhancement decisions.
Predicting Road Network Vulnerability to Fluvial Flooding Using Machine Learning Classifiers: Case Study of Houston during Hurricane Harvey
Abdulla, Bahrulla (author) / Birgisson, Bjorn (author)
Construction Research Congress 2020 ; 2020 ; Tempe, Arizona
2020-11-09
Conference paper
Electronic Resource
English
| British Library Conference Proceedings | 2019