Research And Application Of Uncertain Genetic Neural Network In Landslide Hazard Prediction

China is one of the countries relatively prone geological disasters and landslide occurs particularly frequently,and it will cause a lot of casualties and property losses each year.It is difficult to predict landslide hazard due to the formation conditions and predisposing factor of landslides is extremely complex and uncertain,it has a theoretical and practical significance by taking a reasonable and feasible method to predict the risk of landslide to the prevention and treatment of landslide hazard.Landslide caused by a variety of factors,which is highly nonlinearity and complexity.In order to express the relationship between these factors and the risk of landslides,since the neural network can accurately reflect the complex nonlinear relationship between the parameters within the system,using the sample data to training the neural network,we build a high-resolution genetic neural network model to predict landslide hazard rating,which bring the genetic algorithm to optimize the initial weights and thresholds of back propagation neural network to find global optimum.However,the rainfall that one important factor for the landslide hazard prediction is uncertain attribute whose value is between a range not an exact value,which is difficult to process it effective by the standard back propagation neural network model.Therefore,in order to make better use of the property of uncertain and improve the prediction accuracy of the model,based on the standard genetic neural network,we build an uncertain genetic neural network model by proposing the separation degree of uncertain data for uncertain property rainfall in landslide evaluation,introduced the uncertain data into the classification model,combined with generalized discrete ways to elaborate the process of uncertain data.Eventually,we take Yanan City for example,according to the theory of the study areas of special environmental conditions and local geography geological disasters occurred,combined with the previous research results on this place to selected property,and the selected various properties associated with the landslide was slope pattern,slope height,slope aspect,slope degree,rock and soil mass,rainfall,vegetation and landslide hazard level.The results showed that the method obtain a higher classification accuracy,verifying the feasibility of uncertain genetic neural network in landslide hazard prediction.