Potential Landslide Identification And Module Development Based On Machine Learning

Landslide is a geological disaster with wide distribution,high occurrence frequency and high risk,which brings serious losses to human life and property.Early identification of potential landslides is an effective way to achieve landslide disaster prevention and control.The machine learning method can couple all kinds of landslide influence factors,and quantitatively describe the complex relationship between landslide influence factors and landslides from data.The potential landslide identification model is established to realize the automatic identification of potential landslides.At present,there are the following deficiencies in the study of potential landslide identification based on machine learning :(1)when selecting the machine learning model,it mainly relies on previous studies and subjective experience to select a single and easily lost optimal model;(2)There is a lack of a platform for integrating multiple machine learning models and facilitating process processing,which is not conducive to the engineering application of potential landslide identification.Based on this,this paper studies the potential landslide recognition and module development based on machine learning.The specific work includes :(1)Based on the study of seven machine learning principles and algorithms with different characteristics and performance,such as logistic regression,support vector machine,artificial neural network,naive Bayes,decision tree,random forest and gradient boosting decision tree,a potential landslide recognition module based on machine learning is designed and developed by Python language.This module has the characteristics of clear process,simple interface,certain visual expression,lightweight and installation-free.(2)Taking Ledu District,Haidong City,Qinghai Province,to Minhe Hui Autonomous County and part of the red ancient area of Lanzhou City,Gansu Province as the experimental area,the landslide influencing factors in the experimental area were determined based on the landslide mechanism in the experimental area,and 14 kinds of landslide influencing factors including topography,geology,vegetation cover and human activities were extracted.At the same time,considering that deformation is not only the basic feature of potential landslide,but also the symbol of landslide mechanical instability,the surface deformation obtained based on In SAR technology is also used as a landslide influencing factor in this paper.Then based on the known landslide in the experimental area,combined with slope and deformation rate,landslide and non-landslide samples are extracted.(3)The potential landslide identification experiment in the experimental area is carried out by using the module developed in this paper.The effectiveness of each function of the module is verified,and the potential landslide identification results in the experimental area are obtained,which provides reference for the prevention and control of landslide disasters in the region.The potential landslide identification in the experimental area provides a sample reference for the use of the module,which reflects the advantages of simple and fast operation of the module.The realization of the module is helpful to improve the accuracy and efficiency of potential landslide identification.This study provides a method reference and engineering application platform for potential landslide identification.