Regional landslide susceptibility prediction method based on semi-supervised random forest model

Abstract

The invention relates to a regional landslide susceptibility prediction method based on a semi-supervised random forest model. The method comprises the following steps: S1, carrying out spatial analysis on landslide records and related control factors in a research region, and screening out known landslide samples; s2, based on the frequency ratio and correlation analysis, determining a control factor which can best represent landslide development characteristics, and establishing a random forest model; s3, on the basis of the FR values of the control factors, known landslide grid units and randomly selected non-landslide grid units, performing full-supervised machine learning, namely a random forest model, and according to the five types of landslide susceptibility grades in the step S2, outputting and predicting an initial landslide susceptibility value; s4, expanding the known landslide sample; s5, randomly selecting grid units from the extremely low susceptible areas as non-landslide samples; and S6, establishing a semi-supervised random forest model. According to the invention, the landslide susceptibility prediction modeling performance is further improved on the basis of fully supervised machine learning.

Description

technical field [0001] The invention relates to the technical field of geological disaster prediction, in particular to a method for predicting regional landslide susceptibility based on a semi-supervised random forest model. Background technique [0002] Under the influence of seasonal heavy rainfall and large-scale engineering construction, many landslides occur every year in my country, which often cause serious losses to the safety of local residents, building facilities and the environment. The study of landslide susceptibility can more accurately predict the spatial probability of potential landslide occurrence in a specific area. Therefore, it is necessary to strengthen the research on regional landslide susceptibility prediction to guide the disaster prevention and mitigation work in areas with high landslide incidence. [0003] Machine learning is a multi-disciplinary interdisciplinary major, covering probability theory knowledge, statistical knowledge, approximate...

AI Technical Summary

Problems solved by technology

On the one hand, unsupervised machine learning does not require known samples of landslides and non-landslides as model output variables in the training and testing process, but the lack of guidance of prior knowledge such as landslides and non-landslides makes it difficult to obtain the modeling accuracy of unsupervised machine learning. ensure

On the other hand, landslide susceptibility prediction modeling based on fully supervised machine learning also has some deficiencies, the main manifestations are: 1) It is difficult and expensive to conduct field surveys and obtain landslide sample data

It is generally difficult to obtain all landslide samples in a large-scale research area. It can be seen that the known landslide samples need to be expanded; The testing process brings a large number of errors, which reduces the accuracy of landslide susceptibility prediction

Images