首页|Researchers from Texas A&M University Report Findings in Machine Le arning (Interpretable Machine Learning for Predicting Urban Flash Flood Hotspots Using Intertwined Land and Builtenvironment Features)
Researchers from Texas A&M University Report Findings in Machine Le arning (Interpretable Machine Learning for Predicting Urban Flash Flood Hotspots Using Intertwined Land and Builtenvironment Features)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – Investigators publish new report on Ma chine Learning. According to news reporting originating from College Station, Te xas, by NewsRx correspondents, research stated, “Pluvial flash floods are fast-m oving hazards and causes significant disruptions in urban areas. With the increa se in heavy precipitations, the ability to proactively identify flash floods hot spots in cities is critical for flood nowcasting and predictive monitoring of ri sks.” Financial support for this research came from National Science Foundation (NSF). Our news editors obtained a quote from the research from Texas A&M University, “While rainfall runoff models and hydrologic models are useful model s for flash flood prediction, these models are computationally expensive and eff ort intensive to be used for flood nowcasting. To address this challenge, this s tudy presents interpretable machine learning models for predicting urban flash f lood hotspots based on intertwined land and built environment features. The task of predicting flash flood hotspots is formulated as a binary classification pro blem, and three recent flash flood events in U.S. cities are selected for data c ollection and model validation. Various features related to land and built envir onment characteristics are constructed using diverse datasets, and the occurrenc es of flash floods are captured using crowdsource data from the events. Using th ese features and datasets, the flash flood hotspots of cities are predicted with two ensemble models based on decision trees. The results demonstrate that the m odels can achieve good accuracy (0.8) in identifying flooded/non-flooded locatio ns. Especially, the models can achieve high true positive rate (0.83-0.89) and l ow missing rate, demonstrating the methods’ practicability for accurately predic ting flooded hotspots. The model interpretation results indicate that land featu res related to hydrological and topological features have greater impacts on fla sh flood risk, than built environment features. Further analysis reveals that th e feature importance, model performance, and model transferability performance v ary among cities and localized specifications of the models are needed for accur ate prediction of flash flood for a particular city.”
College StationTexasUnited StatesN orth and Central AmericaCyborgsEmerging TechnologiesMachine LearningTexa s A&M University
NETL
NSTL
