Machine Learning-Based Evaluation of Susceptibility to Geological Hazards in Yunyang District, Shiyan City, China

Regional geohazard susceptibility evaluation and early warning are effective means of disaster prevention and mitigation. The traditional regional geohazard evaluation has problems such as limited model accuracy and insufficient refinement. With the rapid development of big data and artificial intelligence technology, machine learning algorithms are gradually widely used in geologic hazard evaluation and have achieved better results. The paper uses BP neural network model and support vector machine model in machine learning algorithms to predict regional geologic disaster susceptibility. The paper selects Utopia District of Shiyan City, Hubei Province as the study area, constructs the evaluation database, selects the sample set, and trains the evaluation model with tuning parameter optimization. The results show that the support vector machine model has the highest AUC value and the distribution of geologic hazards in the evaluation results is more accurate. The susceptibility of geologic hazards in Utopia is divided into four categories: low susceptibility, medium susceptibility, medium-high susceptibility and high susceptibility, in which the low susceptibility area accounts for 17.11% of the total area, the medium susceptibility area accounts for 33.57% of the total area, the medium-high susceptibility area accounts for 42.94% of the total area and the high susceptibility area accounts for 36.55% of the total area. The results of the thesis research are of guiding significance for the disaster prevention and mitigation work in Shiyan City Utopia. Keywords: geohazard, susceptibility assessment, support vector machine, BP neural network, informativeness modeling