Debris flow represents severe geological hazards in mountainous regions,posing significant risks to human lives and property safety.Accurately predicting their hazard level is paramount.This study utilizes 37 sets of debris flow samples from Yunnan Province as a case study.Initially,the Grey Relational Analysis(GRA)technique is applied to filter and exclude evaluation factors with minimal impact on debris flow hazards,resulting in the identification of 8 core indicators.Subsequently,Principal Component Analysis(PCA)is employed to reduce the dimensionality of these core indicators,extracting principal components.These comprehensive indicators are then fed into an Elman neural network to forecast debris flow hazard levels.The findings reveal that,compared to alternative models,the GRA-PCA-Elman model achieves an accuracy rate of 90.91%and exhibits outstanding generalization capabilities,rendering it suitable for debris flow prediction.GRA effectively eliminates evaluation indicators with relatively minor impacts on debris flow hazards,while PCA efficiently removes correlated information among indicators,thereby enhancing the prediction accuracy of the model.
维普
万方数据