Data-Driven Disaster Management in a Smart City

Abstract

Disasters, both natural and man-made, are complex events that result in the loss of human life and/or the destruction of properties. The advances in Information Technology (IT) and Big Data Analysis represent an opportunity for the development of resilient environments, since from the application of Big Data (BD) technologies it is possible not only to extract patterns of occurrences of events, but also to predict them.The work carried out in this dissertation aims to apply the CRISP-DM methodology to conduct a descriptive and predictive analysis of the events that occurred in the city of Lisbon, with emphasis on the events that affected buildings.Through this research it was verified the existence of temporal and spatial patterns of occurrences with some events occurring in certain periods of the year, such as floods and collapses that are recorded more frequently in periods of high precipitation. The spatial analysis showed that the city center is the area most affected by the occurrences, and it is in these areas where the largest proportion of buildings with major repair needs are concentrated.Finally, machine learning models were applied to the data, and the Random Forest model obtained the best result with an accuracy of 58%. This research contributes to improve the resilience of the city since the analysis developed allowed to extract insights regarding the events and their occurrence patterns that will help the decision-making process.