CN113705607B - Landslide susceptibility evaluation method based on two-step strategy - Google Patents
Landslide susceptibility evaluation method based on two-step strategy Download PDFInfo
- Publication number
- CN113705607B CN113705607B CN202110832986.9A CN202110832986A CN113705607B CN 113705607 B CN113705607 B CN 113705607B CN 202110832986 A CN202110832986 A CN 202110832986A CN 113705607 B CN113705607 B CN 113705607B
- Authority
- CN
- China
- Prior art keywords
- sample
- landslide
- training
- unlabeled
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Pit Excavations, Shoring, Fill Or Stabilisation Of Slopes (AREA)
Abstract
The invention discloses a landslide susceptibility evaluation method based on a two-step strategy, which comprises the following steps: PU-learning (PU-two-step) based on two-step strategy is a learning strategy. The method is actually realized by relying on a base learner, and the support vector machine SVM is selected as the base learner. In sample construction, part of sample points P are randomly extracted from a positive example set P, and added into an unlabeled sample set U to form a new unlabeled sample set U ', the samples in the U' are taken as negative example samples and positive example samples in the positive example set P to form a training data set D, and a naive Bayes classifier is trained; and scoring samples in the unlabeled sample set U by using a trained naive Bayes classifier, setting a threshold value to be 0.5, selecting reliable negative examples, and constructing a reliable negative example sample data set RN. In summary, the invention uses SVM model, uses positive example set P, reliable negative example set RN and Q (namely U-RN) to train iteratively to obtain training model, finally predicts unlabeled sample set U to obtain landslide susceptibility evaluation result.
Description
Technical Field
The invention belongs to the field of engineering geological research, and particularly relates to a landslide vulnerability evaluation method based on a two-step strategy.
Background
With the application of computer technology in various fields in recent years, and the perfection and maturity of a data processing method based on a machine learning theory, many machine learning methods are applied to landslide susceptibility evaluation. In 2012 Che et al analyzed the susceptibility to landslide in volcanic terrain areas around karst Lin Bei (Limbe) using a informative method based on data defining several environmental control factors. Feng Ce by taking a '4.20' reed mountain earthquake region as a research region, taking gradient, fluctuation, land type, fault distance and peak acceleration of earthquake as evaluation indexes based on Remote Sensing (RS) and GIS technologies, constructing an evaluation model by adopting a logistic regression method to evaluate landslide susceptibility indexes of the research region, and checking the effect of the model through a subject working characteristic curve (ROC). Liu Yanfang and the like take the three gorges reservoir region Gu county as a research area, select gradient, elevation, lithology, land utilization, distance from water system and the like as landslide evaluation factors, analyze sensitivity indexes of each influence factor to landslide occurrence by adopting a deterministic coefficient method, and draw an area landslide susceptibility demarcation chart on the basis. Zhang Relin and the like calculate the national landslide susceptibility index by using a probability ratio method and draw a national susceptibility zoning map. Wu Xueling in order to solve the problems that the traditional landslide prediction data source is limited, the data updating period is long, the rule hidden in a complex landslide system is difficult to find easily, and the like, three gorges reservoir area is taken as a research object, data of each landslide influence factor is extracted from multi-source space data, a slope unit is divided in an area by adopting a digital terrain hydrologic analysis method, the selected environmental influence factors are resampled, two types of support vector machine models are built, quantitative relations between various environmental influence factors and occurrence of landslide are analyzed, and an area landslide susceptibility partition map is generated. Jusheng and the like, by using an artificial neural network model (ANN) and a logistic regression model (LR), landslide susceptibility evaluation is carried out on the Yuan' an county of Hubei province, a regional landslide susceptibility demarcation diagram is obtained, and a comparison analysis is carried out on the prediction result. The research shows that 9 types of index factors related to landslide occurrence are extracted according to experience knowledge at the position of the district co-development historical landslide 177, elevation factors are removed by means of correlation analysis, the rest 8 types of factors are finally selected for landslide susceptibility evaluation, and the landslide susceptibility regional map of the research district is obtained by means of ArcGIS and SPSS model software analysis and calculation. Xia Hui by taking Wushan county in Sanxia reservoir as a research area, extracting 9 environmental influence factors (elevation, gradient, slope direction, terrain humidity index TWI, surface roughness index TRI, stratum lithology, water system distance, construction distance and vegetation coverage index NDVI) through analysis, removing elevation factors through factor correlation analysis, and analyzing disaster points and environmental factors through a Support Vector Machine (SVM) and an Artificial Neural Network (ANN) model to obtain a landslide susceptibility regional map of the research area. Pal et al used an extreme learning machine, predicted landslide susceptibility indexes for areas near england litter baud town using two remote sensing datasets (one multispectral and one hyperspectral), and mapped landslide susceptibility risk plots.
Since loess soil is soft, it is very vulnerable to water and soil loss, and thus various geological disasters frequently occur on loess plateau for a long period of time. Nowadays, with the increasing of human economy and engineering activities, under the dual influence of human activities and natural transitions, the types and the number of geological disasters are increasing, wherein landslide disasters have the trend of increasing occurrence frequency and occurrence scale. The special landform environment of the loess plateau provides conditions and foundation for the generation and inoculation of loess landslide.
In recent years, geological disasters frequently occur on loess plateau, and the control work on the geological disasters is increasingly receiving attention from society in face of more and more severe disaster situations. In order to safely and effectively plan and build towns, the disaster prevention and reduction work of the cities is developed, and the regional landslide disaster susceptibility regional map becomes an important reference factor.
Disclosure of Invention
The invention aims to overcome the defects, and provides a landslide hazard susceptibility evaluation method based on a two-step strategy, which is characterized in that the contribution degree of each factor to landslide occurrence and inoculation is analyzed by a qualitative or quantitative method, and finally landslide hazard susceptibility evaluation analysis is carried out on a research area by comparing and selecting an evaluation model or method suitable for disaster characteristics of the research area.
In order to achieve the above object, the present invention comprises the steps of:
s1, acquiring a plurality of environmental factors from a data source according to loess causes, standardizing a grid graph of all the environmental factors into a grid unit set, extracting a plurality of sample data from the grid unit set to serve as an unlabeled data set U, and taking data in a history landslide inventory as a positive example set P;
s2, constructing a reliable negative example sample data set by using a spy method;
s3, training by adopting a base learner to obtain a training model;
s4, predicting each sample in the unlabeled data set U by adopting a training model, determining the category of the unlabeled point according to a threshold value, and finally obtaining a landslide susceptibility evaluation result.
In S1, the environmental factors include elevation, slope direction, curvature, normalized vegetation index NDVI, geologic structure, loess erosion intensity, and terrain humidity index TWI.
In S1, the grid graph of the environmental factor is normalized to a 30m×30m grid cell set.
The specific method of S2 is as follows:
randomly extracting part of sample points P from the positive example set P, adding the sample points P into an unlabeled data set U to form a new unlabeled sample set U ', taking samples in the new unlabeled sample set U' as negative example samples, forming a training data set D by the negative example samples and the positive example samples in the positive example set P, training, scoring the samples in the unlabeled sample set U by adopting a trained model, selecting negative example samples meeting the requirements, and constructing a reliable negative example sample data set RN.
The training adopts a naive Bayes classifier, and the trained model is the trained naive Bayes classifier.
And S3, performing iterative training on the set of the positive example set P, the reliable negative example sample data set RN and the untagged data set U by adopting an SVM model, and removing the set of the reliable negative example sample data set RN to obtain a training model.
Compared with the prior art, the method has the advantages that the unlabeled data set and the positive case set are extracted according to loess causes, the reliable negative case sample data set is constructed by adopting a spy method, the training is carried out by adopting the base learner to obtain a training model, each sample in the unlabeled data set U is predicted by adopting the training model, the category of the unlabeled point is determined according to the threshold value, and finally the landslide susceptibility evaluation result is obtained.
Compared with the prior art, the invention has the following beneficial effects:
according to loess causes, the contribution degree of each environmental factor to landslide generation and inoculation is analyzed, and 8 relevant environmental factors are given out; constructing a reliable negative example sample data set by utilizing a spy method; training by adopting a basic learner to obtain a training model. It can be seen that the invention does not require a large number of known landslide sample points to build a positive sample data set, so that the robustness of the model is significantly improved. According to the landslide susceptibility evaluation method, a training model is adopted to predict each sample in the unlabeled data set U, the category of the unlabeled point is determined according to the threshold value, and finally the landslide susceptibility evaluation result is obtained. Therefore, the model can be more stable by randomly selecting the negative example sample and repeatedly training the base learner for a plurality of times, and the prediction precision is also obviously improved.
Drawings
FIG. 1 is a grid plot of historical landslide points and elevations in an embodiment;
FIG. 2 is a diagram of historical landslide points and grade grids in an embodiment;
FIG. 3 is a diagram of historical landslide points and a sloping grid in an embodiment;
FIG. 4 is a grid graph of historical landslide points and curvatures in an embodiment;
FIG. 5 is a grid of historical landslide points and vegetation cover indices in an example;
FIG. 6 is a grid view of historical landslide points and geologic structures in an embodiment;
FIG. 7 is a grid plot of historical landslide points and yellow map erosion intensities in an embodiment;
FIG. 8 is a grid plot of historical landslide and terrain wetness indexes in an embodiment;
FIG. 9 is a schematic diagram of the PU-two-step method-SVM landslide easy-to-go region of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention comprises the following steps:
1. constructing an original data set;
according to the research on loess landslide causes in a research area and past experience and research, 8 environmental factors are obtained from a data source: elevation, grade, slope direction, curvature, normalized vegetation index (NDVI), geologic structure, loess erosion intensity, topography humidity index (TWI) as the main environmental factors affecting the development of loess landslide in the research area, normalizing the grid pattern of these 8 environmental factors into a 30m×30m grid unit set, and extracting about 180 ten thousand sample data of the whole area therefrom as an unlabeled data set U; 203 history landslide in the history landslide catalog are taken as a positive example set P.
2. Constructing a reliable negative example sample data set by using a spy method;
the training data set is constructed by utilizing a Spy (Spy method), namely, part of sample points P are randomly extracted from a positive example set P and added into an unlabeled sample set U to form a new unlabeled sample set U ', the samples in the U' are used as negative example samples and positive example samples in the positive example set P to form a training data set D, a naive Bayesian classifier is trained, the trained naive Bayesian classifier is utilized to score the samples in the unlabeled sample set U, a threshold value is set to be 0.5, reliable negative example samples are selected, and a reliable negative example sample data set RN is constructed.
The principle of the naive Bayes classifier is to calculate the posterior probability according to the prior probability of the sample, namely the probability that the sample belongs to a certain class, and finally select the class with the maximum posterior probability as the class to which the sample belongs. The specific description of the algorithm is as follows:
a) Each sample uses an n-dimensional eigenvector x= { X 1 ,x 2 ,L,x n And } represents. Feature vectors are used to describe n attributes A 1 ,A 2 ,Λ,A n N dimensions of the sample.
b) Assuming n classes, respectively using C 1 ,C 2 ,Λ,C n And (3) representing. Given an unknown sample X, if the naive Bayes classifier assigns the unknown sample X to C i Class, the following formula must be satisfied:
P(C i |X i )>P(C j |X j )(1≤j≤n,j≠i)； (1)
wherein P (C) i I X) is the maximum posterior probability.
c) At C i In the case where the prior probability of the class is unknown, P (C 1 )＝P(C 2 )＝Λ＝P(C n ) Otherwise P (C i )＝s i And/s. Wherein s is i Is C i The number of training samples in the class, s, is the total number of training samples. As can be seen from the bayesian formula, in order to calculate P (C i I X), P (x|c) needs to be calculated i )P(C i ) And P (X|C) i ) May be derived from the training dataset.
3. Training by a basic learner;
and (3) utilizing an SVM model, and utilizing a positive example set P, a reliable negative example set RN and Q (namely U-RN) to carry out iterative training to obtain a training model.
The invention selects a support vector machine SVM as a base learner. The basic principle of the Support Vector Machine (SVM) is to find an optimal classification hyperplane in a sample space, so that the distance from the nearest point to the classification hyperplane in two types of sample points to the hyperplane is the largest. The larger the distance, the better the generalization of the classifier and the lower the error. Given data set d= { (x) 1 ,y 1 ),(x 2 ,y 2 ),K,(x N ,y N ) I=1, 2, …, N, x i For the ith feature vector, y i Is a class label. Given a hyperplane: wx (Wx) i +b=0, then the geometric spacing of the hyperplane with respect to the sample point is:
the minimum value of the geometric spacing of the hyperplane with respect to all sample points is:
i.e. the distance of the support vector to the hyperplane. Order theSolving the maximum split hyperplane problem for the SVM model is expressed as:
this is a convex quadratic programming problem with inequality constraints, introducing a Lagrangian function:
wherein alpha is i Is Lagrangian multiplier, and alpha i And is more than or equal to 0. The deviation is obtained by:
substituting the binary problem into the Lagrangian function to obtain the dual problem of the original planning problem:
the model can be obtained after solving:
in the case of linear separable, the support vector machine model is a linear classifier. In the case of linear inseparable, the support vector machine only considers points near the boundary line, i.e. points far from the boundary, when calculating, and has no effect on the model establishment. This greatly reduces the amount of computation, and introduces Lagrangian pairs when the objective function is established, enabling the mapping computation of complex kernel functions.
4. Predicting sample points in unlabeled data sets to obtain landslide susceptibility evaluation results;
and predicting each sample in the unlabeled data set U by using a training model, and determining the category of the unlabeled point according to a threshold value so as to obtain a landslide susceptibility evaluation result.
The invention combines the geological environment of the research area on the basis of the development and distribution characteristics of the historical landslide points in the collection area, screens out the most important disaster-causing factors from a plurality of environmental factors capable of inducing landslide disasters, analyzes the contribution degree of each factor to landslide generation and inoculation by using a qualitative or quantitative method, and finally carries out landslide disaster susceptibility evaluation analysis on the research area by comparing and selecting an evaluation model or method suitable for the disaster characteristics of the research area.
Examples:
basic geological data and historical landslide cataloging data of a research area are collected by taking Lingtai county in Pinggan, gansu province as the research area, appropriate environmental factors are selected on the basis of a landslide generation mechanism on a loess plateau where the research area is located and previous research experience, eight environmental factors of the research area are divided into grid units with the same size by using ArcGIS drawing software, the data of each environmental factor are extracted from the grid units, and landslide disaster susceptibility of the research area is evaluated.
Collecting 30m resolution Digital Elevation Model (DEM) data of a research area for extracting topography and landform information; 1500000 geologic map for extracting geologic structure information; a normalized vegetation index map of 8 months in 2017 for extracting vegetation coverage information; a loess etching map in 2017 for extracting loess etching intensity; landslide cataloging and field data investigation are used for acquiring historical landslide data of a research area. According to an expert evaluation method, 8 environmental factors are selected: elevation, grade, slope, curvature, normalized vegetation index (NDVI), geologic structure, loess erosion intensity, and terrain humidity index (TWI) as influencing factors, as shown in fig. 1-8.
According to the characteristics of the research area, selecting regular 30m×30m grid units to divide the research area into about 180 ten thousand grid units, extracting data from the 8 environmental factors collected in the previous step, and further normalizing. Because of the particularities of the geological industry, field researchers are generally interested in only historical landslide, and therefore only historical landslide records of the research area and about 180 ten thousand unknown units are collected in the data. It follows that this is not a simple two-classification problem, but a problem of how to train a model from a large number of unlabeled samples and a small number of known landslide samples, and then find a sample with a high likelihood of landslide from the large number of unlabeled samples.
The landslide susceptibility of the research area is evaluated by using a model based on a two-step strategy PU-learning (PU-two-step method-SVM), and the result is shown in figure 9, wherein most of the historical landslide points in the landslide susceptibility area map of the Lingtai county obtained by using the PU-two-step method-SVM model fall in the area with high landslide susceptibility predicted by the model, and are consistent with the actual historical landslide data.
Claims (3)
1. The landslide susceptibility evaluation method based on the two-step strategy is characterized by comprising the following steps of:
s1, acquiring a plurality of environmental factors from a data source according to loess causes, standardizing a grid graph of all the environmental factors into a grid unit set, extracting a plurality of sample data from the grid unit set to serve as an unlabeled data set U, and taking data in a history landslide inventory as a positive example set P;
the environmental factors comprise elevation, gradient, slope direction, curvature, normalized vegetation index NDVI, geological structure, loess erosion intensity and topography humidity index TWI;
the grid graph of the environmental factors is normalized to a 30m x 30m grid cell set;
s2, constructing a reliable negative example sample data set by using a spy method, wherein the specific method comprises the following steps of:
randomly extracting part of sample points P from the positive example set P, adding the sample points P into an unlabeled data set U to form a new unlabeled sample set U ', taking samples in the new unlabeled sample set U' as negative example samples, forming a training data set D by the negative example samples and the positive example samples in the positive example set P, training, scoring the samples in the unlabeled sample set U by adopting a trained model, selecting negative example samples meeting the requirements, and constructing a reliable negative example sample data set RN;
s3, training by adopting a base learner to obtain a training model;
the training method comprises the steps of utilizing an SVM model, and utilizing a positive example set P, a reliable negative example set RN and Q (namely U-RN) to carry out iterative training to obtain a training model;
the basic principle of the support vector machine SVM is that an optimal classification hyperplane is found in a sample space, so that separation in two types of sample points is realizedThe closest point of the hyperplane-like surface is the largest in distance to the hyperplane, the greater the distance, the better the generalization of the classifier, the lower the error, and the given dataset D = { (x) 1 ,y 1 ),(x 2 ,y 2 ),...,(x N ,y N ) I=1, 2, …, N, x i For the ith feature vector, y i Is a class mark; given a hyperplane: wx (Wx) i +b=0, then the geometric spacing of the hyperplane with respect to the sample point is:
the minimum value of the geometric spacing of the hyperplane with respect to all sample points is:
i.e., the distance of the support vector to the hyperplane, letSolving the maximum split hyperplane problem for the SVM model is expressed as:
s.t.y i (w·x i +b)≥1,i＝1,2,...,N，
this is a convex quadratic programming problem with inequality constraints, introducing a Lagrangian function:
wherein alpha is i Is Lagrangian multiplier, and alpha i And (3) not less than 0, and obtaining the deviation guide of the above formula:
substituting the binary problem into the Lagrangian function to obtain the dual problem of the original planning problem:
αi≥0,i＝1,2,...,m,
the model can be obtained after solving:
s4, predicting each sample in the unlabeled data set U by adopting a training model, determining the category of the unlabeled point according to a threshold value, and finally obtaining a landslide susceptibility evaluation result.
2. The landslide vulnerability assessment method based on two-step strategy of claim 1, wherein the training adopts a naive Bayesian classifier, and the trained model is the trained naive Bayesian classifier.
3. The landslide vulnerability evaluation method based on the two-step strategy according to claim 1, wherein in S3, an SVM model is adopted to perform iterative training on a set of positive example set P, reliable negative example sample data set RN and unlabeled data set U excluding the reliable negative example sample data set RN, so as to obtain a training model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110832986.9A CN113705607B (en) | 2021-07-22 | 2021-07-22 | Landslide susceptibility evaluation method based on two-step strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110832986.9A CN113705607B (en) | 2021-07-22 | 2021-07-22 | Landslide susceptibility evaluation method based on two-step strategy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113705607A CN113705607A (en) | 2021-11-26 |
| CN113705607B true CN113705607B (en) | 2023-08-22 |
Family
ID=78650316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110832986.9A Active CN113705607B (en) | 2021-07-22 | 2021-07-22 | Landslide susceptibility evaluation method based on two-step strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705607B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109359738A (en) * | 2018-10-19 | 2019-02-19 | 西南交通大学 | A kind of Landslide hazard appraisal procedure based on QPSO-BP neural network |
| CN111582386A (en) * | 2020-05-11 | 2020-08-25 | 四川师范大学 | Random forest based geological disaster multi-disaster comprehensive risk evaluation method |
| CN111898315A (en) * | 2020-07-27 | 2020-11-06 | 中国科学院空天信息创新研究院 | Landslide susceptibility assessment method based on fractal-machine learning hybrid model |
| WO2020249125A1 (en) * | 2019-06-14 | 2020-12-17 | 第四范式（北京）技术有限公司 | Method and system for automatically training machine learning model |
| CN112966722A (en) * | 2021-02-07 | 2021-06-15 | 南昌大学 | Regional landslide susceptibility prediction method based on semi-supervised random forest model |
-
2021
- 2021-07-22 CN CN202110832986.9A patent/CN113705607B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109359738A (en) * | 2018-10-19 | 2019-02-19 | 西南交通大学 | A kind of Landslide hazard appraisal procedure based on QPSO-BP neural network |
| WO2020249125A1 (en) * | 2019-06-14 | 2020-12-17 | 第四范式（北京）技术有限公司 | Method and system for automatically training machine learning model |
| CN111582386A (en) * | 2020-05-11 | 2020-08-25 | 四川师范大学 | Random forest based geological disaster multi-disaster comprehensive risk evaluation method |
| CN111898315A (en) * | 2020-07-27 | 2020-11-06 | 中国科学院空天信息创新研究院 | Landslide susceptibility assessment method based on fractal-machine learning hybrid model |
| CN112966722A (en) * | 2021-02-07 | 2021-06-15 | 南昌大学 | Regional landslide susceptibility prediction method based on semi-supervised random forest model |
Non-Patent Citations (1)
| Title |
|---|
| 邱维蓉 ; 吴帮玉 ; 潘学树 ; 唐亚明 ; .几种聚类优化的机器学习方法在灵台县滑坡易发性评价中的应用.西北地质.2020,(第01期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113705607A (en) | 2021-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A hybrid model considering spatial heterogeneity for landslide susceptibility mapping in Zhejiang Province, China | |
| Chen et al. | Performance evaluation of GIS-based new ensemble data mining techniques of adaptive neuro-fuzzy inference system (ANFIS) with genetic algorithm (GA), differential evolution (DE), and particle swarm optimization (PSO) for landslide spatial modelling | |
| Lv et al. | A hybrid ensemble-based deep-learning framework for landslide susceptibility mapping | |
| Wu et al. | Landslide susceptibility assessment using object mapping units, decision tree, and support vector machine models in the Three Gorges of China | |
| Zhao et al. | A comparative study of landslide susceptibility mapping using SVM and PSO-SVM models based on Grid and Slope Units | |
| Yang et al. | An integrative hierarchical stepwise sampling strategy for spatial sampling and its application in digital soil mapping | |
| Wan et al. | A knowledge-based decision support system to analyze the debris-flow problems at Chen-Yu-Lan River, Taiwan | |
| CN111898315B (en) | Landslide susceptibility assessment method based on fractal-machine learning hybrid model | |
| Zhang et al. | A review of machine learning based Species' distribution modelling | |
| CN113902580B (en) | Historical farmland distribution reconstruction method based on random forest model | |
| Niu et al. | Susceptibility assessment of landslides triggered by the Lushan earthquake, April 20, 2013, China | |
| Yu et al. | A landslide susceptibility map based on spatial scale segmentation: A case study at Zigui-Badong in the Three Gorges Reservoir Area, China | |
| CN114298436A (en) | Landslide susceptibility evaluation method based on Bagging strategy | |
| Mao et al. | Landslide hazards mapping using uncertain Naïve Bayesian classification method | |
| Gao et al. | Application of convolutional neural network fused with machine learning modeling framework for geospatial comparative analysis of landslide susceptibility | |
| Xiong et al. | Landslide susceptibility mapping using ant colony optimization strategy and deep belief network in Jiuzhaigou Region | |
| Xia et al. | Land resource use classification using deep learning in ecological remote sensing images | |
| Venkatesan et al. | An improved Bayesian classification data mining method for early warning landslide susceptibility model using GIS | |
| Mao et al. | Landslide susceptibility modelling based on AHC-OLID clustering algorithm | |
| Zhao et al. | An improved spatial case-based reasoning considering multiple spatial drivers of geographic events and its application in landslide susceptibility mapping | |
| Zhao et al. | Mapping landslide sensitivity based on machine learning: A case study in Ankang City, Shaanxi Province, China | |
| Ghosh et al. | Meta-transfer learning: An application to streamflow modeling in river-streams | |
| CN117408167A (en) | Debris flow disaster vulnerability prediction method based on deep neural network | |
| CN113705607B (en) | Landslide susceptibility evaluation method based on two-step strategy | |
| CN115907091A (en) | Earthquake staff death assessment method based on PSO-SVR |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |