Landslide susceptibility evaluation method based on fractal-machine learning hybrid model

Abstract

The invention discloses a landslide susceptibility evaluation method based on a fractal-machine learning hybrid model. The method comprises the following steps: selecting landslide susceptibility evaluation factors; analyzing a fractal relationship between historical landslide geological disaster points and landslide susceptibility evaluation factors in the experimental area based on a fractal model, and calculating a preliminary landslide susceptibility index on the basis of solving fractal dimensions between the landslide susceptibility evaluation factors and the historical geological disaster points; constructing a multi-scene sample data set: constructing sample data sets of three different scenes by the three non-landslide samples and a unified landslide sample; taking the sample datasets of the three scenes respectively as inputs of the NB model and the SVM model to carry out landslide susceptibility evaluation research. Compared with a negative sample generated from a low-slopearea and a non-landslide area in traditional landslide susceptibility research, the negative sample quantitatively selected based on the fractal model can improve the quality of a landslide susceptibility evaluation sample, and the use of the fractal-machine learning hybrid model can improve the accuracy of landslide susceptibility evaluation.

Description

technical field [0001] The invention relates to the technical field of landslide susceptibility assessment of machine learning models. Specifically, it is a landslide susceptibility assessment method based on a fractal-machine learning hybrid model. Background technique [0002] In landslide susceptibility assessment studies based on machine learning models, the selection of positive and negative samples is an important aspect that affects the prediction performance of landslide susceptibility assessment models and the accuracy of landslide susceptibility assessment results. In traditional landslide susceptibility studies, negative samples are generated from low-slope areas and no-landslide areas: negative samples based on slope information will cause the final susceptibility assessment results to rely too much on slope single information, and weaken other geographic— The impact of environmental factors on landslides. When the accuracy of the slope map used is low, it will...

AI Technical Summary

Problems solved by technology

In traditional landslide susceptibility studies, negative samples are generated from low-slope areas and no-landslide areas: negative samples based on slope information will cause the final susceptibility assessment results to rely too much on slope single information, and weaken other geographic— Effects of Environmental Factors on Landslides

When the accuracy of the slope map used is low, it will directly lead to the obvious clustering of non-landslide samples generated in the low-slope area in space, and a serious unbalanced distribution phenomenon will appear, which will eventually affect the generalization ability of the machine learning model; rather than landslides The generation method of the samples is purely qualitative. Only according to the occurrence of landslides in the current study area, the areas that have not yet occurred landslides are identified as landslide-free areas, and these areas have the possibility of landslide geological disasters in the future stage, and there have never been landslides. The method of generating negative samples completely ignores the influence of all geographical-environmental factors including slope information on landslide geological hazards.

Images