Landslide susceptibility prediction model based on principal component analysis and extreme learning machine

Abstract

The invention relates to the technical field of geological disaster prediction, in particular to a landslide susceptibility prediction model based on principal component analysis and an extreme learning machine, and the model comprises the following steps: S1, obtaining landslide catalog and landslide susceptibility modeling related environmental factors of a research area; s2, carrying out dimensionality reduction on the environmental factors by utilizing principal component analysis, calculating a principal component score as an initial landslide susceptibility value, and dividing different susceptibility intervals; s3, superposing the extremely high incidence area and the remote sensing image, determining a landslide hidden danger point as an expanded landslide sample through visual interpretation, and forming a landslide sample by the landslide record and the expanded landslide sample; s4, grid units are randomly selected from the extremely low susceptible areas to serve as non-landslide samples; and S5, establishing an extreme learning machine prediction model. The correlation between environmental factors and repeated information reflected by the environmental factors during comprehensive evaluation can be eliminated. The redundancy of the data subjected to dimension reduction is greatly reduced through principal component analysis, and time is saved for subsequent calculation.

Description

technical field [0001] The invention relates to the technical field of geological disaster prediction, in particular to a landslide susceptibility prediction model based on principal component analysis and extreme learning machine. Background technique [0002] As a common geological disaster, landslides seriously threaten the lives and property safety of our people. In landslide susceptibility prediction modeling, a large number of environmental factors are usually selected, and there is a strong correlation between some environmental factors, which makes the model affected by repeated information. On the other hand, landslide catalogs are usually used as landslide samples, but it is difficult and incomplete to obtain landslide sample data in the field; when selecting non-landslide samples, they are usually randomly selected from the research area outside the landslide point, lacking prior knowledge of landslides and non-landslides. With the guidance of knowledge, random s...

AI Technical Summary

Problems solved by technology

On the other hand, landslide catalogs are usually used as landslide samples, but it is difficult and incomplete to obtain landslide sample data in the field; when selecting non-landslide samples, they are usually randomly selected from the research area outside the landslide point, lacking prior knowledge of landslides and non-landslides. With the guidance of knowledge, random selection of non-landslide samples in the entire study area will cause a lot of errors, and the accuracy of modeling is difficult to guarantee

Images