earticle

영어

In this study, geospatial data big data reflecting the physical and environmental status of buildings and satellite image data was constructed, and disaster risk factors were analyzed by constructing a disaster prediction model based on it. Big data of spatial information were generated, from which the physical and environmental status of structures and data of satellite images were reflected. Big data of urban space were generated based on 230 regions in South Korea for developing a building disaster prediction model. The generated big data of spatial information as well as the history of structural damages to structures due to torrential rain, strong wind, heavy snow, and typhoons from 2008 to 2017 were utilized in creating the disaster prediction model. The collected variables were applied in a random forest model, which are machine learning approaches, to develop the structure disaster prediction model. Looking at the importance of variables, in heavy rain, it was found that the precipitation intensity, semi-underground buildings, the threshold of the heavy rain, the frequency of occurrence of the threshold of the heavy rain, and the status of underground buildings had the greatest influence on the damage to buildings due to heavy rain. In the case of strong winds, the higher the normalized city index, steep slope, and poor drainage, the greater the influence. In heavy snow, the number of days of new snow exceeding 20 cm, hazardous materials treatment and storage facilities, average height, and old buildings had the greatest impact on disasters.