Landslide susceptibility prediction method based on machine learning and evaluation unit combinatorial optimization

Abstract

The invention relates to the technical field of geological disaster prediction and prevention, in particular to a landslide susceptibility prediction method based on machine learning and evaluation unit combinatorial optimization, comprising the following steps: acquiring environmental geological data including a plurality of evaluation factors in a research area; respectively extracting a grid unit and a slope unit as evaluation units based on the environmental geological data; combining a random forest RF model, a limit gradient boost XGBoost model and the evaluation unit to form a plurality of prediction combination models; the prediction combination model is evaluated according to the prediction combination model performance indexes, and an optimal combination model is determined; and generating a landslide susceptibility prediction result corresponding to the research area based on the optimal combination. The prediction precision is improved while the calculation efficiency is improved, and the universality of the prediction method is enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of geological disaster prediction and prevention, and in particular relates to a landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units. Background Art

[0002] Landslides are a widespread and devastating type of geological disaster in nature. According to statistics, landslides cause annual economic losses exceeding tens of billions of dollars worldwide and result in numerous casualties. Scientific and accurate prediction of landslide susceptibility is a key prerequisite for disaster prevention and control. Its core lies in establishing a quantitative relationship model between landslide probability and environmental geological conditions.

[0003] There are many limitations in the existing technologies for predicting landslide susceptibility. 1. The evaluation methods are simplistic. Knowledge-driven methods such as the analytic hierarchy process and expert scoring methods rely on subjective experience and lack objectivity and replicability. 2. The physical models constructed have poor versatility. For example, the limit equilibrium method requires precise geotechnical parameters and is only applicable to small areas, making it difficult to extend to complex geological environments. 3. The choice of evaluation units is controversial. For example, grid units are widely used due to their ease of operation, but their regular shapes ignore terrain continuity and make it difficult to accurately characterize landslide-prone environments. Although slope units are more in line with terrain characteristics, the division process is complex, computational efficiency is low, and there is a lack of standardized extraction processes. 4. The model's generalization ability is insufficient. Traditional machine learning models such as logistic regression and support vector machines have limited ability to fit high-dimensional nonlinear data and lack systematic comparison and optimization, resulting in limited prediction accuracy. Therefore, how to improve prediction accuracy while improving computational efficiency and enhancing the versatility of prediction methods has become an urgent problem to be solved. Summary of the Invention

[0004] In view of this, the present invention aims to provide a landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units, which improves the prediction accuracy while improving the computational efficiency and the versatility of the prediction method.

[0005] To achieve the above object, the technical solution created by the present invention is implemented as follows: The present invention provides a landslide susceptibility prediction method based on machine learning and combined optimization of evaluation units, comprising the following steps: acquiring environmental geological data of a study area including multiple evaluation factors; extracting grid units and slope units as evaluation units based on the environmental geological data; utilizing a random forest RF model and an extreme gradient boosting XGBoost model in combination with the evaluation units to form a plurality of prediction combination models; evaluating the prediction combination models according to performance indicators of the prediction combination models to determine an optimal combination model; and generating a landslide susceptibility prediction result corresponding to the study area based on the optimal combination model.

[0006] Furthermore, after obtaining environmental geological data of the study area including multiple evaluation factors, the method includes: performing Pearson correlation analysis on the evaluation factors; if the correlation coefficient of any two of the evaluation factors exceeds 0.7, then eliminating the evaluation factor with the lowest contribution among the any two evaluation factors.

[0007] Furthermore, the multiple evaluation factors include DEM data, and grid units and slope units are extracted as evaluation units based on the environmental geological data, specifically including: filling depressions on the DEM data to generate a depression-free topographic map; calculating the direction and cumulative amount of water flow on the depression-free topographic map to extract the river network; dividing the river network into preliminary slope units based on a watershed algorithm, and correcting the boundaries of the slope units; converting the corrected slope units into vector surface data as the final slope units.

[0008] Furthermore, a random forest RF model, an extreme gradient boosting XGBoost model and the evaluation unit are combined to form a plurality of prediction combination models, specifically including: constructing the RF model and the XGBoost model; and dynamically optimizing the hyperparameters of the prediction combination model through a network search method.

[0009] Furthermore, after dynamically optimizing the hyperparameters of the prediction combination model through a network search method, the method further includes: constructing a landslide sample set, marking historical landslide points as positive samples and non-landslide points as negative samples; dividing the landslide sample set into a training set and a test set; and using the training set to train the prediction combination model.

[0010] Furthermore, the predictive combination model is evaluated according to the predictive combination model performance index to determine the optimal combination, specifically including: using the test set to verify the predictive combination model performance index of the predictive combination model; and selecting the predictive combination model with the highest area under the receiver operating characteristic curve AUC value in the predictive combination model performance index as the optimal combination model.

[0011] Furthermore, based on the optimal combination model, a landslide susceptibility prediction result corresponding to the study area is generated, specifically including: dividing the landslide susceptibility test results into susceptibility zones including low, medium, high and extremely high susceptibility zones based on the natural discontinuity method; and verifying the rationality of the susceptibility zones according to the landslide density.

[0012] Furthermore, after generating the landslide susceptibility prediction results corresponding to the study area based on the optimal combination model, the method also includes: performing importance analysis on the multiple evaluation factors based on the Gini coefficient and average gain, selecting dominant factors from the multiple evaluation factors for ranking; superimposing the dominant factors with the susceptibility zones to generate a prevention and control recommendation map, in which the fault zones, steep slope areas and high rainfall areas that need priority reinforcement are marked.

[0013] In addition, the present invention also provides a landslide susceptibility prediction system based on machine learning and evaluation unit combination optimization, including: an environmental geological data acquisition module, used to obtain environmental geological data of the study area including multiple evaluation factors; an evaluation unit extraction module, used to extract grid units and slope units as evaluation units based on the environmental geological data; a prediction combination generation module, used to use the random forest RF model, the extreme gradient boosting XGBoost model and the evaluation unit combination to form a variety of prediction combination models; a prediction combination evaluation module, used to evaluate the prediction combination model according to the prediction combination model performance index, and determine the optimal combination model; a landslide susceptibility prediction result generation module, used to generate a landslide susceptibility prediction result corresponding to the study area based on the optimal combination model.

[0014] In addition, the present invention also provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the landslide susceptibility prediction method described above.

[0015] Compared to existing technologies, the present invention provides a landslide susceptibility prediction method based on machine learning and combined optimization of evaluation units. This method acquires environmental geological data for a study area, including multiple evaluation factors. Based on this environmental geological data, grid units and slope units are extracted as evaluation units. Multiple prediction combinations are formed using the random forest (RF) model and the extreme gradient boosting (XGBoost) model combined with the evaluation units. The prediction combination models are evaluated based on their performance indicators to determine the optimal combination model. Based on the optimal combination model, landslide susceptibility prediction results for the corresponding study area are generated. This method improves computational efficiency while enhancing prediction accuracy and enhancing the versatility of the prediction method. A comprehensive evaluation using performance metrics such as the area under the receiver operating characteristic curve (AUC), accuracy (Acc), and recall (Recall) is performed to select the optimal prediction combination. Furthermore, by optimizing hyperparameters, eliminating redundant factors, and integrating the topographical characterization capabilities of slope units, prediction accuracy is significantly improved, overcoming the limitations of traditional single models and unit combinations and providing a scientific basis for landslide disaster prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: Figure 1 A schematic flow chart of a landslide susceptibility prediction method provided by an embodiment of the present invention; Figure 2 A schematic diagram of the Pearson correlation coefficient of evaluation factors of grid cells in the landslide susceptibility prediction method provided by an embodiment of the present invention; Figure 3 A schematic diagram of the Pearson correlation coefficient of the evaluation factor of the slope unit in the landslide susceptibility prediction method provided by an embodiment of the present invention; Figure 4 A schematic diagram showing the visualization of the AUC values of four prediction combinations in the landslide susceptibility prediction method provided in an embodiment of the present invention; Figure 5 A schematic diagram of accuracy verification indicators for prediction combinations in the landslide susceptibility prediction method provided by an embodiment of the present invention; Figure 6 A radar chart of the dominant factors in the landslide susceptibility prediction method provided by an embodiment of the present invention; Figure 7 A schematic diagram of the structure of an electronic device provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation to the present invention. Similar elements in different embodiments use associated similar element numbers. In the following embodiments, many detailed descriptions are intended to enable the present invention to be better understood. However, those skilled in the art can easily recognize that some of the features can be omitted in different situations, or can be replaced by other elements, materials, or methods. In some cases, some operations related to the present invention are not shown or described in the specification. This is to avoid the core part of the present invention being overwhelmed by too much description. For those skilled in the art, it is not necessary to describe these related operations in detail. They can fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0018] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other to form various implementation methods. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various orders in the description and the drawings are only for the purpose of clearly describing a certain embodiment and are not intended to be a required order, unless otherwise specified that a certain order must be followed.

[0019] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0020] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art can understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0021] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0022] Example 1 like Figure 1As shown, an embodiment of the present invention provides a landslide susceptibility prediction method based on machine learning and evaluation unit combination optimization, comprising the following steps: S10: acquiring environmental geological data of a study area including multiple evaluation factors; S20: extracting grid units and slope units as evaluation units based on the environmental geological data; S30: combining a random forest RF model, an extreme gradient boosting XGBoost model and the evaluation unit to form a plurality of prediction combination models; S40: evaluating the prediction combination model according to the performance index of the prediction combination model to determine the optimal combination model; S50: generating a landslide susceptibility prediction result for the corresponding study area based on the optimal combination model.

[0023] In order to solve the problems of lack of unified standards for the selection of evaluation models and evaluation units and limited prediction accuracy in existing landslide susceptibility prediction methods, the landslide susceptibility prediction method based on machine learning and evaluation unit combination optimization proposed in the embodiment of the present invention combines two learning models, random forest (RF) and XGBoost algorithm, and two evaluation units, grid unit and slope unit, to construct at least four prediction combination models: RF+grid unit, RF+slope unit, XGBoost+grid unit, XGBoost+slope unit. The performance indicators such as the area under the receiver operating characteristic curve (AUC), accuracy (Acc), and recall rate (Recall) are used to comprehensively evaluate the performance of the prediction combination model to screen out the optimal combination model, such as Figure 2 and Figure 3 This embodiment of the present invention overcomes the limitations of the traditional single learning model and evaluation unit combination, providing a scientific basis for landslide disaster prevention and control. Similarly, it can also be applied to regional landslide risk assessment, land use planning, infrastructure site selection, and disaster emergency management.

[0024] This method builds a high-precision prediction combination model by integrating the Random Forest (RF) and XGBoost algorithms with two spatial evaluation units: grid unit and slope unit, realizing a combined architecture of multiple models and multiple evaluation units. The evaluation unit can be divided into: Grid unit: 30m×30m regular grid, covering an area of 2449.15km2 of the study area ² , generating a total of 3,491,938 units; slope units: extracted based on hydrological analysis method, combined with high-resolution images (resolution ≤ 2m) to manually correct boundaries, and finally generating 5,537 units.

[0025] Machine learning model selection: Two algorithms, random forest (RF) and XGBoost, were integrated and combined with grid cells and slope cells respectively to form four prediction combination models (RF+grid cell, RF+slope cell, XGBoost+grid cell, and XGBoost+slope cell).

[0026] Furthermore, in the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units provided in an embodiment of the present invention, environmental geological data of the study area including multiple evaluation factors is obtained. The data source of the environmental geological data can be: Topographic data: Geospatial data cloud platform GDEMV3 (30m resolution DEM); Remote sensing data: Landsat8 OLI image (NDVI calculation); Geological data: National Geological Information Center (lithology classification, fault distribution vector map); Meteorological data: National Earth System Science Data Center (monthly rainfall from 1901 to 2023, extracted annual means).

[0027] The extraction method of evaluation factors can be: Terrain evaluation factors: Elevation, slope, and aspect: calculated based on DEM using ArcGIS spatial analysis tools; Profile curvature and plane curvature: generated using the Curvature tool to reflect the terrain's unevenness and water flow rate.

[0028] Geological evaluation factors: Lithology type: vectorize geological maps and convert them into raster data; Distance from fault: Euclidean distance analysis, generating a buffer raster.

[0029] Meteorological and hydrological evaluation factors: Average annual rainfall: spatial interpolation (kriging) to generate rainfall distribution maps; TWI (Topographic Wetness Index): Based on the formula Calculate, where A is the catchment area and β is the slope; SPI (Hydrodynamic Index): The formula is , reflecting the runoff erosion capacity.

[0030] Human activity evaluation factors: Distance to roads / water systems: Generates a Euclidean distance raster based on road and water system vector data.

[0031] Furthermore, the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units provided in an embodiment of the present invention, after obtaining the environmental geological data of the study area including multiple evaluation factors, also includes: performing Pearson correlation analysis on the evaluation factors; if the correlation coefficient of any two of the evaluation factors exceeds 0.7, then eliminating the evaluation factors with low contribution among any two.

[0032] Preliminary evaluation factors include: terrain evaluation factors (elevation, slope, aspect, profile curvature, plan curvature), geological evaluation factors (lithology, distance from faults), meteorological evaluation factors (average annual rainfall), Normalized Difference Vegetation Index (NDVI), Topographic Wetness Index (TWI), Standardized Precipitation Index (SPI), and human activity evaluation factors (distance from roads and water systems), totaling 13 factors.

[0033] Factor screening: Redundant evaluation factors were eliminated through Pearson correlation analysis (threshold 0.7). For example, when the correlation coefficient between TWI and SPI reached 0.78, TWI was retained and SPI was eliminated. In order to ensure the independence and objectivity of each factor and avoid redundant information from interfering with each other, which would reduce the prediction performance of the model, the Pearson correlation test was used to measure the correlation between factors using landslide disaster points as sample data to further complete the evaluation factor screening. Figure 3 As shown in the figure, when grid cells are used as evaluation units, most pairwise correlation coefficients between factors are less than 0.4. The highest correlation coefficient between elevation and rainfall is 0.65, but this is still below the moderate correlation threshold of 0.7, indicating that the correlations between factors are not significant. All factors meet the independence requirement and can be used to build models and make predictions. When slope cells are used as evaluation units, most pairwise correlation coefficients between factors are less than 0.7. However, it is noted that the correlation coefficient between TWI and SPI reaches 0.78 (absolute value). After removing one of the two and quickly building a comparative model, the model prediction performance is better when TWI is retained than when SPI is retained, so SPI is removed.

[0034] Furthermore, in the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units provided by an embodiment of the present invention, multiple evaluation factors include DEM data, and grid units and slope units are extracted as evaluation units based on environmental geological data, specifically including: filling depressions on the DEM data to generate a depression-free topographic map; calculating the water flow direction and water flow accumulation on the depression-free topographic map to extract the river network; dividing the river network into preliminary slope units based on the watershed algorithm, and correcting the boundaries of the slope units; converting the corrected slope units into vector surface data as the final slope units.

[0035] Grid cells can be extracted by creating regular 30m×30m grid cells in ArcGIS, covering the entire study area, and calculating the mean value (continuous variables) or mode (categorical variables) of each evaluation factor in each grid cell.

[0036] Slope unit extraction can be done by filling depressions in the DEM data to eliminate local depressions and generate a depression-free topographic map to ensure the continuity of water flow direction. Flow direction and flow rate are then calculated on this depression-free topographic map. The D8 algorithm is used to determine the flow direction, and areas with cumulative flow rate ≥ 1000 pixels are marked as river networks. Watershed-based division is performed using the watershed algorithm to divide the river network into preliminary slope units, using the river network as the boundary. Slope unit boundaries can be manually modified by adjusting the slope unit boundaries using high-resolution imagery (such as Google Earth) to ensure the consistency of the terrain within the slope unit. Fragmentary units with an area less than 0.1 km² are eliminated, and adjacent small slope units are merged. Finally, 5,537 slope units are generated and converted into vector surface data as the final slope units.

[0037] Furthermore, in the landslide susceptibility prediction method based on machine learning and evaluation unit combination optimization provided by an embodiment of the present invention, a random forest RF model, an extreme gradient boosting XGBoost model and an evaluation unit are combined to form a variety of prediction combination models, specifically including: constructing an RF model and an XGBoost model; and dynamically optimizing the hyperparameters of the prediction combination model through a network search method.

[0038] It's important to note that the accuracy of the constructed forecast combination model depends not only on the quality of the data itself but also on the hyperparameter settings, which also determine the model's training method and forecast structure. Using accuracy as the objective function, multiple rounds of grid search were conducted to determine the optimal hyperparameter settings for each forecast combination model.

[0039] The first step is to build a forecast combination model: Hyperparameter web search: RF: max_depth (5-15), n_estimators (50-200), min_samples_split (2-10); XGBoost: learning_rate (0.01-0.2), max_depth (3-8), reg_alpha (0.1-1.0).

[0040] The grid search method combined with five-fold cross validation is used to achieve dynamic optimization of hyperparameters and balance the fitting ability and generalization performance of the prediction combination model. Five-fold cross validation can prevent overfitting and improve the robustness of the prediction combination model.

[0041] Then, the hyperparameters of the prediction combination model are dynamically optimized: Random Forest (RF) model algorithm principle: multiple decision trees are generated through Bootstrap sampling, and the final result is determined by voting. Loss function: Gini Index; the formula is: ,in is the proportion of samples in category i.

[0042] Dynamically optimized hyperparameters: prediction combination with grid cells: max_depth=10, n_estimators=80, min_samples_split=7; prediction combination with slope cells: max_depth=10, n_estimators=150, min_samples_split=4.

[0043] Then comes the XGBoost model. The algorithm principle is based on the gradient boosting framework, which reduces the loss function by iteratively optimizing the decision tree. Objective function: ,in, is the loss function and Ω is the regularization term.

[0044] Hyperparameter optimization: prediction combination with grid cells: learning_rate=0.03, max_depth=6, reg_alpha=0.5; prediction combination with slope cells: learning_rate=0.1, max_depth=5, reg_alpha=0.3.

[0045] It can be seen that the landslide susceptibility prediction method based on machine learning and evaluation unit combination optimization proposed in the embodiment of the present invention significantly improves the prediction accuracy and the versatility of the prediction method through the comparison of multiple prediction combination models, redundancy control of evaluation factors, refined extraction of slope units and hyperparameter optimization.

[0046] Furthermore, the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units proposed in an embodiment of the present invention, after dynamically optimizing the hyperparameters of the prediction combination through a network search method, also includes: constructing a landslide sample set, marking historical landslide points as positive samples, and marking non-landslide points as negative samples; dividing the landslide sample set into a training set and a test set; and using the training set to train the prediction combination model.

[0047] Positive samples: Based on historical landslide inventories, remote sensing interpretation, and field surveys, a total of 115 landslide coordinates were obtained; each landslide point corresponds to an evaluation unit (grid unit or slope unit) and is marked as 1. Negative samples: Non-landslide points were randomly selected at a 1:1 ratio, preferentially meeting the following conditions: flat areas with a slope of less than 5°; areas with high vegetation cover with an NDVI greater than 0.6; or stable areas greater than 500 m from the fault. These points were marked as 0 to ensure sample balance.

[0048] Dataset division: training set: 70% samples (161), used for training the prediction combination model; test set: 30% samples (69), used for verifying the prediction combination model.

[0049] Furthermore, in the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units proposed in an embodiment of the present invention, the prediction combination is evaluated based on model performance indicators to determine the optimal combination model. Specifically, the model performance indicators are verified using a test set; and the prediction combination model with the highest area under the receiver operating characteristic (ROC) curve (AUC) value is selected as the optimal combination model. Performance indicators may include AUC, accuracy (Acc), recall, and F1-score.

[0050] Verification and result analysis of the forecast combination model: Calculation of performance indicators of the forecast combination model: AUC: Area under the ROC curve, which measures the overall discrimination ability of the prediction combination model; Accuracy (Acc): ; Recall: ; F1-Score: ; Based on the confusion matrix, the area under the receiver operating characteristic curve (ROC) AUC value is often used to evaluate the prediction performance of the model, such as Figure 4 As shown. The AUC value range is [0, 1]. The closer the AUC is to 1, the better the classifier performance and the higher the accuracy of the model. The ROC curve is plotted with the true positive rate TPR (y-axis) and the false positive rate FPR (x-axis) as the axes. Based on this, the accuracy rate Accuracy (Acc), recall rate Recall and precision Precision are also commonly used to verify the accuracy of the model, such as Figure 5 As shown in the formula: TP is the true positive, representing the class correctly classified as landslide; FP is the false positive, representing the class incorrectly classified as landslide; FN is the false negative, representing the class incorrectly classified as non-landslide; TN is the true negative, representing the class correctly classified as non-landslide.

[0051] The performance comparison of the forecast combination model is shown in Table 1.

[0052] Table 1

[0053] Furthermore, in the landslide susceptibility prediction method based on the combined optimization of machine learning and evaluation units proposed in an embodiment of the present invention, a landslide susceptibility prediction result for the corresponding study area is generated based on the optimal combination, specifically including: dividing the landslide susceptibility test results into susceptibility zones including low, medium, high and extremely high susceptibility zones based on the natural discontinuity method; and verifying the rationality of the susceptibility zones according to the landslide density.

[0054] Based on the natural break method, the predicted probability is divided into four levels of susceptibility zones: low, medium, high, and very high. Susceptibility zoning rules: Based on the natural break method, the predicted probability is divided into four levels: low susceptibility zone (0-0.25), medium susceptibility zone (0.25-0.50), high susceptibility zone (0.50-0.75), and very high susceptibility zone (0.75-1.00). Spatial distribution characteristics: The extremely high susceptibility areas are concentrated in fault zones (<200 m from the fault), slopes >25°, and areas with an average annual rainfall >650 mm; The low susceptibility area is distributed in the western plains (slope <10°, NDVI>0.6).

[0055] Verify the rationality of susceptibility zoning based on landslide density: Table 2

[0056] Furthermore, the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units proposed in an embodiment of the present invention, after generating the landslide susceptibility prediction results of the corresponding study area based on the optimal combination model, also includes: performing importance analysis on multiple evaluation factors based on the Gini coefficient and average gain, selecting dominant factors from multiple evaluation factors for ranking, and quantifying the contribution of the dominant factors; superimposing the dominant factors with the susceptibility zones to generate a prevention and control recommendation map, in which the fault zones, steep slope areas and high rainfall areas that need priority reinforcement are marked.

[0057] In the random forest model, the contribution of an evaluation factor to classification purity is quantified based on the reduction in the Gini coefficient. In the XGBoost model, the degree to which an evaluation factor reduces the loss function is measured based on the average gain (Split Gain). The ranking of the leading factors is shown in Table 3.

[0058] Table 3

[0059] The susceptibility zones are superimposed with the dominant factors (rainfall, fault zones, steep slope zones), especially the extremely high susceptibility zones are superimposed with the susceptibility zones, to generate a prevention and control recommendation map. The prevention and control recommendation map marks the fault zones, steep slope zones and high rainfall areas that need priority reinforcement, such as road-cut slopes and areas around residential areas.

[0060] Using the landslide susceptibility prediction method based on machine learning and combined optimization of evaluation units provided by the embodiment of the present invention, the optimal combined model (RF + slope unit) achieved an AUC of 0.92, significantly improving the prediction accuracy achieved by traditional methods by 10%-15%. Computational efficiency and accuracy are balanced: it can be seen that grid units are suitable for rapid assessment, while slope units are suitable for scenarios requiring high precision. Enhanced interpretability and practicality: The ranking of the importance of dominant factors and the generation of prevention and control recommendation maps provide a direct basis for engineering decision-making, see [1]. Figure 6 The universality and scalability of the landslide susceptibility prediction method are improved: the framework formed by this landslide susceptibility prediction method can be extended to the prediction of other geological disasters such as debris flow and collapse.

[0061] Example 2 In addition, the present invention also provides a landslide susceptibility prediction system based on machine learning and evaluation unit combination optimization, including: an environmental geological data acquisition module, used to obtain environmental geological data of the study area including multiple evaluation factors; an evaluation unit extraction module, used to extract grid units and slope units as evaluation units based on the environmental geological data; a prediction combination model generation module, used to use the random forest RF model, the extreme gradient boosting XGBoost model and the evaluation unit combination to form a variety of prediction combination models; a prediction combination model evaluation module, used to evaluate the prediction combination model according to the model performance index and determine the optimal combination model; a landslide susceptibility prediction result generation module, used to generate the landslide susceptibility prediction result of the corresponding study area based on the optimal combination model.

[0062] To address the lack of unified standards for selecting evaluation models and evaluation units in existing landslide susceptibility prediction methods, which limits prediction accuracy, the present invention proposes a landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units. This method combines two learning models, the random forest (RF) and XGBoost algorithms, with two evaluation units, the grid unit and the slope unit, to construct at least four prediction combination models: RF + grid unit, RF + slope unit, XGBoost + grid unit, and XGBoost + slope unit. The performance of the prediction combination models is comprehensively evaluated using performance metrics such as the area under the receiver operating characteristic curve (AUC), accuracy (Acc), and recall (Recall), to select the optimal combination model. This embodiment of the present invention addresses the limitations of traditional single learning model and evaluation unit combinations, providing a scientific basis for landslide disaster prevention and control. Similarly, the method can also be applied to regional landslide risk assessment, land use planning, infrastructure site selection, and disaster emergency management.

[0063] This method builds a high-precision prediction combination model by integrating the Random Forest (RF) and XGBoost algorithms with two spatial evaluation units: grid unit and slope unit, realizing a combined architecture of multiple models and multiple evaluation units. The evaluation unit can be divided into: Grid unit: 30m×30m regular grid, covering an area of 2449.15km2 of the study area ² , generating a total of 3,491,938 units; slope units: extracted based on hydrological analysis method, combined with high-resolution images (resolution ≤ 2m) to manually correct boundaries, and finally generating 5,537 units.

[0064] Machine learning model selection: Two algorithms, random forest (RF) and XGBoost, were integrated and combined with grid cells and slope cells respectively to form four prediction combination models (RF+grid cell, RF+slope cell, XGBoost+grid cell, and XGBoost+slope cell).

[0065] In addition, embodiments of the present invention provide an electronic device, a readable storage medium, and a computer program product. These include a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is in operation, the processor and the memory communicate via the bus. The machine-readable instructions are executed by the processor to execute the steps of the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units, as described above. Figure 7 A computer device, a readable storage medium, and a computer program product are provided in the embodiments of the present invention.

[0066] Figure 7 FIG. 1 is a structural diagram of a computer device 12 provided in an embodiment of the present invention. Figure 7 A block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention is shown. Figure 7 The computer device 12 shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0067] like Figure 7 As shown, computer device 12 is represented in the form of a general-purpose computing device. Computer device 12 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0068] Components of computer device 12 may include, but are not limited to, one or more processors or processing units 16 , a system memory 28 , and a bus 18 that connects various system components, including system memory 28 and processing unit 16 .

[0069] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0070] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0071] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be configured to read and write non-removable, non-volatile magnetic media ( Figure 7 Not shown, often called a "hard drive"). Although Figure 7 Not shown, a magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention.

[0072] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program modules 42 generally implement the functions and / or methods of the embodiments described herein.

[0073] The computer device 12 may also communicate with one or more external devices 14 (e.g., a keyboard, a pointing device, a display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may occur through an input / output (I / O) interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 20. Figure 7 As shown, network adapter 20 communicates with the other modules of computer device 12 via bus 18. It should be understood that although not shown in the figures, other hardware and / or software modules may be used in conjunction with computer device 12, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0074] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units provided in an embodiment of the present invention.

[0075] An embodiment of the present invention also provides a non-transitory computer-readable storage medium storing computer instructions, on which a computer program is stored, wherein when the program is executed by a processor, the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units provided in all the inventive embodiments of this application is realized.

[0076] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device.

[0077] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0078] The program code that comprises on the computer-readable medium can be transmitted with any appropriate medium, includes but not limited to wireless, electric wire, optical cable, RF etc., or above-mentioned any suitable combination.Can write the computer program code that is used to carry out the operation of the present invention with one or more programming languages or its combination, described programming language comprises object-oriented programming language such as Java, Smalltalk, C++, also comprises conventional procedural programming language--such as " C " language or similar programming language.Program code can be carried out on user's computer completely, partly on user's computer, carry out as an independent software package, partly on user's computer partly on remote computer, or carry out completely on remote computer or server.In the situation that relates to remote computer, remote computer can comprise local area network (LAN) or wide area network (WAN) to be connected to user's computer by the network of any kind, perhaps, can be connected to external computer (for example, utilize Internet service provider to come to connect by Internet).

[0079] An embodiment of the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the above-mentioned landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units.

[0080] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0081] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units, characterized by: The following steps are involved: Obtain environmental geological data of the study area including multiple evaluation factors; Extracting grid units and slope units as evaluation units based on the environmental geological data; Utilize the random forest RF model, the extreme gradient boosting XGBoost model and the evaluation unit to form a variety of prediction combination models; Evaluate the forecast combination model according to the forecast combination model performance index to determine the optimal combination model; A landslide susceptibility prediction result corresponding to the study area is generated based on the optimal combination model.

2. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 1 is characterized in that: After obtaining environmental geological data of the study area including multiple evaluation factors, the method includes: Performing Pearson correlation analysis on the evaluation factors; If the correlation coefficient between any two of the evaluation factors exceeds 0.7, the evaluation factor with the lowest contribution among the two evaluation factors is eliminated.

3. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 1 or 2, characterized in that: The multiple evaluation factors include DEM data, and grid units and slope units are extracted as evaluation units based on the environmental geological data, specifically including: Performing depression-filling processing on the DEM data to generate a depression-free topographic map; Calculating the direction of water flow and the cumulative amount of water flow on the non-depression topographic map to extract the river network; Dividing the river network into preliminary slope units based on a watershed algorithm, and revising the boundaries of the slope units; The modified slope unit is converted into vector surface data as the final slope unit.

4. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 3 is characterized by: The random forest RF model, the extreme gradient boosting XGBoost model and the evaluation unit are combined to form a variety of prediction combination models, including: Constructing the RF model and the XGBoost model; The hyperparameters of the prediction combination model are dynamically optimized through a network search method.

5. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 4 is characterized in that: After dynamically optimizing the hyperparameters of the forecast combination model through a network search method, the method further includes: Construct a landslide sample set, mark historical landslide points as positive samples, and non-landslide points as negative samples; Dividing the landslide sample set into a training set and a test set; The prediction combination model is trained using the training set.

6. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 5 is characterized in that: Evaluate the forecast combination model based on the forecast combination model performance indicators to determine the optimal combination, specifically including: Using the test set to verify the prediction combination model performance index of the prediction combination model; The prediction combination model with the highest area under the receiver operating characteristic curve (AUC) value among the prediction combination model performance indicators is selected as the optimal combination model.

7. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 1 is characterized in that: The landslide susceptibility prediction results corresponding to the study area are generated based on the optimal combination model, specifically including: Dividing the landslide susceptibility test results into susceptibility zones including low, medium, high and very high susceptibility zones based on the natural break method; The rationality of the susceptibility zoning was verified based on landslide density.

8. The landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units according to claim 7 is characterized in that: After generating a landslide susceptibility prediction result corresponding to the study area based on the optimal combination model, the method further includes: performing importance analysis on the plurality of evaluation factors based on the Gini coefficient and the average gain, and selecting dominant factors from the plurality of evaluation factors for ranking; The dominant factors are superimposed on the susceptibility zones to generate a prevention and control recommendation map, in which fault zones, steep slope areas and high rainfall areas that require priority reinforcement are marked.

9. A landslide susceptibility prediction system based on combined optimization of machine learning and evaluation units, characterized by: include: Environmental geological data acquisition module, used to obtain environmental geological data of the study area including multiple evaluation factors; An evaluation unit extraction module, configured to extract grid units and slope units as evaluation units based on the environmental geological data; A prediction combination generation module is used to combine the random forest RF model and the extreme gradient boosting XGBoost model with the evaluation unit to form multiple prediction combination models; A forecast combination evaluation module is used to evaluate the forecast combination model according to the forecast combination model performance index and determine the optimal combination model; The landslide susceptibility prediction result generating module is used to generate the landslide susceptibility prediction result corresponding to the study area based on the optimal combination model.

10. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the processor is running, the machine-readable instructions execute the steps of the landslide susceptibility prediction method based on combined optimization of machine learning and evaluation units as described in any one of claims 1 to 8.

Images