Enhancing Urban Resilience to Flooding in Hydrogeological Risk Areas Through Big Data Analytics Using Deep Neuro-Fuzzy System

Abstract

Urban areas worldwide are increasingly at risk from hydrogeological hazards, leading to severe consequences. Urban flooding and mismanagement of water resources, resulting in riverine flooding, are primary contributors to this risk. Utilizing big data, including mobile phone signals collected at high frequencies, alongside administrative data, is essential for developing risk exposure indicators in smaller urban regions. Accurately assessing human traffic flows and movements is crucial for mitigating the impacts of natural disasters and ensuring a high quality of life in smart cities. However, comprehensive solutions to these challenges are lacking in many countries. Therefore, this study focuses on analyzing the impact of traffic data flow analysis in hydrogeological risk areas. The study employs mobile phone signals as big data to analyze traffic flows and forecast exposure risks to aid decision-making. To ensure data reliability, a circle search integrated fully connected conditional neural network (CS-ConNN) is used for data cleaning, categorizing mobile phone signal data into normal, empty, and garbage. Additionally, the study uses a deep recurrent neuro fuzzy system (DRNFS) to analyze the compound seasonality of circulation flow data and forecast risks, providing alerts to individuals transiting through affected areas. The model is validated through a case study of "Mandolossa," and developed area prone to inundating near Brescia, using hourly data from September 2020 to August 2021. Experimental results and cross-validation demonstrate a forecasting accuracy of 98.975%.