Back propagation (BP) neural network landslide forecasting method based on optimized input layer

Abstract

The invention discloses a back propagation (BP) neural network landslide forecasting method based on an optimized input layer. The BP neural network landslide forecasting method includes steps: (1) related parameters of landslide of the input layer are collected; (2) BP neural network training is performed; (3) whether requirements are met is judged, the known expected value is t, errors are calculated according to an error formula of an output node, sample training is completed if the requirements are met, otherwise Vij theta i is firstly adjusted and then Wij theta j is adjusted according to the error reducing principle, and the corrected value is used for training until the requirements are met; and (4) after sample training is completed, the initialized forecasting sample value is input, and the forecasting value can be obtained. The BP neural network is adopted so that the BP neural network landslide forecasting method is simple to operate, strong in learning ability and good in adaptability.

Description

technical field [0001] The invention relates to the field of natural disaster monitoring and forecasting, in particular to a method for forecasting landslides based on a BP neural network of an optimized input layer. Background technique [0002] At present, the methods for landslide risk prediction can be divided into three categories: qualitative evaluation, quantitative evaluation, and uncertainty analysis. Qualitative evaluation methods include historical analysis and engineering geological analogy. Quantitative evaluation methods include rigid body limit equilibrium method and stress-strain analysis method. Among them, the stress-strain analysis method can be divided into: continuous medium mechanics method and discontinuous medium mechanics method. Uncertain analysis methods include reliability method, fuzzy method, gray system theory, cluster analysis, genetic algorithm and so on. [0003] The first two types of methods above all need to give a certain model for ca...

AI Technical Summary

Problems solved by technology

Even the existing method of using BP neural network to predict landslides has low disaster prediction accuracy due to its narrow selection of the input layer.

Images