Landslide disaster early warning method and system based on stacked auto-encoders
By adopting a stacked autoencoder-based method in landslide disaster warning, and using fuzzy logic and stacked meta-model to optimize the landslide early warning model, the problem of existing models relying on historical data and lack of self-regulation is solved, and the accuracy of landslide displacement prediction is improved and reliability.
Patent Information
- Application Number
- CN202411072441.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-12-06
AI Technical Summary
The existing landslide displacement prediction model relies on a large amount of historical data, lacks self-regulation ability, and has model deviations, which affects the prediction accuracy.
The landslide disaster warning method based on stacked autoencoder is adopted, and the landslide monitoring data is obtained for preprocessing, and the landslide early warning model is optimized by using a fuzzy layer based on fuzzy logic, a linear layer that processes linear transformation, and a decision-making layer based on stacked meta-model. Implement self-regulation and optimization of the model.
Effectively process and identify abnormal data patterns, reduce model deviations, and improve the accuracy and reliability of predicted outputs.
Smart Images

Figure CN119091583A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of disaster warning, and in particular to a landslide disaster warning method and system based on stacked autoencoders. Background Art
[0002] With the continuous progress of the economy, more and more engineering activities have begun to cause varying degrees of damage to natural slopes, so the possibility of landslide disasters is increasing, and there is a trend of increasing severity. The blasting activities in coal mining will cause the rock and soil of the slope to vibrate, and then break and cause landslides. Landslide disasters are one of the most common geological disasters in mines. Once they occur, they may cause serious safety accidents. Therefore, by simulating the changing trend of landslide displacement and timely warning of dangerous displacement, it can help mine managers better understand the displacement of the slope, take effective preventive measures in time, and avoid accidents.
[0003] In the existing field of landslide displacement prediction, various neural network models, such as graph convolutional networks (GCN), recurrent neural networks (RNN) and support vector regression (SVR), have been proven to have wide application value. The above models are good at processing complex nonlinear relationships, but their effective training requires a large amount of historical data and requires precise parameter tuning in order to accurately capture and predict the various factors affecting landslide displacement. In addition, the above models lack the ability to self-regulate new data and have model bias, which in turn affects the final output. Summary of the invention
[0004] One of the embodiments of this specification provides a landslide disaster early warning method based on a stacked autoencoder. The landslide disaster early warning method includes:
[0005] S1: Obtain landslide monitoring data of the target mountain;
[0006] S2: preprocessing the landslide monitoring data to obtain landslide characteristic data and landslide displacement threshold;
[0007] S3: inputting the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained based on the optimization training of the initial landslide early warning model, and is composed of a fuzzy layer based on fuzzy logic, a linear layer for processing linear transformation, and a decision layer constructed based on a stacked metamodel;
[0008] S4: Determine a landslide warning result based on the predicted value of the landslide displacement.
[0009] One of the embodiments of this specification provides a landslide disaster early warning system based on a stacked autoencoder, comprising:
[0010] An acquisition module is used to acquire landslide monitoring data of a target mountain;
[0011] A preprocessing module, used for preprocessing the landslide monitoring data to obtain landslide characteristic data and landslide displacement threshold;
[0012] A prediction module, used for inputting the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained based on the optimization training of the initial landslide early warning model, and is composed of a fuzzy layer based on fuzzy logic, a linear layer for processing linear transformation, and a decision layer constructed based on a stacked metamodel;
[0013] The determination module is used to determine the landslide warning result based on the predicted value of the landslide displacement.
[0014] One of the embodiments of the present specification provides a landslide disaster warning device based on a stacked autoencoder, including a processor, wherein the processor is used to execute a landslide disaster warning method based on a stacked autoencoder.
[0015] One of the embodiments of this specification provides a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a landslide disaster warning method based on a stacked autoencoder.
[0016] The beneficial effects of the present invention are:
[0017] (1) Applying autoencoders to landslide disaster warning tasks, using their powerful feature encoding capabilities to automatically extract key information from large amounts of data and effectively process and identify abnormal data patterns;
[0018] (2) By stacking meta-models in the encoder, the model can be self-regulated and optimized, reducing model bias and improving the final output of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:
[0020] Figure 1 is an exemplary flow chart of a landslide disaster early warning method based on a stacked autoencoder according to some embodiments of this specification;
[0021] Figure 2 is an exemplary flow chart of landslide early warning model processing according to some embodiments of this specification;
[0022] Figure 3is an exemplary flow chart of a landslide disaster early warning system based on a stacked autoencoder according to some embodiments of this specification;
[0023] Figure 4 This is a comparison chart of evaluation parameters of the landslide disaster early warning method based on the stacked autoencoder according to an embodiment of the present invention and the original autoencoder on the test set;
[0024] Figure 5 A loss graph for each round of training of the landslide disaster early warning method based on the original autoencoder according to an embodiment of the present invention;
[0025] Figure 6 This is a loss graph for each round of training of the landslide disaster early warning method based on stacked autoencoders described in an embodiment of the present invention. DETAILED DESCRIPTION
[0026] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0027] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.
[0028] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0029] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.
[0030] Figure 1 FIG. 1 is an exemplary flow chart of a landslide disaster early warning method based on a stacked autoencoder according to some embodiments of this specification. Figure 2 As shown, the landslide disaster early warning method includes the following steps.
[0031] Step S1, obtaining landslide monitoring data of a target mountain.
[0032] The target mountain refers to the mountain where mining, engineering construction and other activities are required. The landslide monitoring data of the target mountain includes the landslide displacement, rainfall, effective horizontal stress, vertical displacement of monitoring points and other geological and environmental data related to the landslide, and a landslide data set is constructed. The landslide monitoring data can be collected and obtained through various landslide detection sensors, or directly imported and obtained through system files.
[0033] Step S2: pre-process the landslide monitoring data to obtain landslide characteristic data and a landslide displacement threshold, wherein the landslide displacement threshold refers to a standard value used to determine whether the observed value in the landslide detection data is abnormal.
[0034] Specifically, step S2 further includes step S21: determining the outliers in the landslide data set based on the Z-score method and step S22: determining the landslide displacement anomaly threshold by the quartile method.
[0035] Step S21: Determine the outliers in the landslide data set based on the Z-score method. Specifically, it also includes: calculating the Z-score of each observation value in the landslide data set; judging whether the observation value is an outlier based on the absolute value of the Z-score; and selecting the Z-score normalization method to process the outliers based on the distribution and characteristics of all observation value data in the data set.
[0036] In some embodiments, the Z score is used to check whether there are outliers in the data, and the outliers are processed using scaling and normalization methods. First, for each value in the data set, its Z score is calculated. The Z score represents the difference between the value and the mean, measured in units of standard deviation. The calculation formula of the Z score is:
[0037]
[0038] Where X is the value in the dataset, μ is the mean, and σ is the standard deviation.
[0039] Then, the absolute value of the Z score is used to determine whether the observation is an outlier. When the absolute value of the Z score is greater than 3, the value is marked as an outlier and recorded. Finally, based on the distribution and characteristics of the data, the Z score normalization method is selected to handle outliers.
[0040] Step S22: Determine the abnormal threshold of landslide displacement by quartile method, which specifically includes: for each observation value in the landslide data set, calculate the lower quartile Q1 and upper quartile Q3 of the observation value; regard the data points outside the range of [Q1-1.5*IQR, Q3+1.5*IQR] as abnormal values.
[0041] In some embodiments, the average value, standard deviation, minimum value, median value, maximum value, lower quartile, upper quartile and other data characteristics of the statistical data are used to determine the abnormal threshold of landslide displacement using the quartile method. First, the lower quartile (Q1) and upper quartile (Q3) of the data are calculated. Then, the data points outside the range of [Q1-1.5*IQR, Q3+1.5*IQR] are regarded as outliers, where IQR (interquartile range) is the difference between Q3 and Q1.
[0042] S3: Inputting the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained based on the optimization training of the initial landslide early warning model, and is composed of a fuzzy layer based on fuzzy logic, a linear layer for processing linear transformation, and a decision layer constructed based on a stacked metamodel.
[0043] In some embodiments, a metamodel layer can be stacked in the encoder FNN. The metamodel layer can self-regulate and optimize the model by learning deeper feature representations from the output of the FNN. The FNN first processes the data based on fuzzy logic and uses Gaussian functions to calculate the membership of the input features. Subsequently, the FNN converts the output of the fuzzy layer into a feature space that is easy for the model to understand. Finally, the metamodel acts as a decision layer to further process the feature space and optimize the received data features.
[0044] For detailed steps on landslide warning model processing, please refer to Figure 2 and related embodiments thereof.
[0045] S4: Determine the landslide warning result based on the predicted value of the landslide displacement. Output the predicted result of the landslide displacement to complete the model prediction and warning.
[0046] It should be noted that the above description of the landslide disaster early warning method flow is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the landslide disaster early warning method flow under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.
[0047] Figure 2is an exemplary flow chart of landslide warning model processing according to some embodiments of this specification, wherein the landslide warning model is an autoencoder composed of an FNN encoder and a GRU decoder, and the processing process includes the following steps:
[0048] S31: Input the processed landslide feature data set into the fuzzy layer and use the Gaussian function to calculate the membership of the input feature. FNN first processes the data based on fuzzy logic and uses the Gaussian function to calculate the membership of the input feature. For each fuzzy rule, the membership is calculated as follows:
[0049]
[0050] Among them, R i is the fuzzy rule parameter vector, x is the input data vector, γ(x) i is the membership degree of the input data x to the i-th rule;
[0051] S32: The membership of the output of the fuzzy layer is input into the linear layer for transformation. FNN converts the output of the fuzzy layer into a feature space that is easy for the model to understand. The membership is transformed through the linear layer:
[0052] f(x)=ω·γ(x)+b,
[0053] Among them, ω is the weight matrix, γ(x) i is the membership vector output by the fuzzy layer, and f(x) is the feature space output after linear transformation.
[0054] S33: Use the ReLU activation function to increase nonlinearity:
[0055] ReLU(f(x)) i =max(0,f(x) i ),
[0056] Among them, f(x) i is the i-th element of the linear layer output vector, and the ReLU function sets all negative values to 0.
[0057] S34: Input the output of the linear layer into the decision layer constructed based on the stacked meta-model. The encoder uses the meta-model as the decision layer to further process the feature space and optimize the received data features:
[0058] M=W i ReLU(f(x)) i +B i ,
[0059] Among them, W i is the weight matrix of the meta-model layer, B iis the bias vector of the meta-model layer, M is the predicted value output by the encoder, and the predicted value refers to the value predicted by the data in the historical landslide data value based on the landslide early warning model.
[0060] In some embodiments, the landslide warning model is obtained based on the optimization training of the initial landslide warning model, wherein the optimization training process of the landslide warning model includes: obtaining a historical landslide data set, wherein the historical landslide data set includes multiple groups of historical landslide displacement feature data; inputting the historical landslide data set into the initial landslide warning model to obtain a training prediction value, wherein the initial landslide warning model is an autoencoder constructed based on a stacked metamodel, and after feature extraction by a deep feature optimizer, the feature space of the upper layer transmission is further refined; based on the training prediction value and the true value input into the evaluation module, the evaluation index is determined, and the evaluation index includes MAE, MSE and RMSE.
[0061] Specifically, a historical landslide training data set is obtained for preprocessing, wherein the historical landslide training set includes multiple groups of historical landslide displacement data.
[0062] In some embodiments, the historical landslide training data set refers to a historical publicly available landslide displacement data set, which includes geological and environmental data related to mine landslides, such as landslide body displacement, rainfall, effective horizontal stress, and vertical displacement of monitoring points; for example, the number of historical landslide training data sets is 23,630 sets.
[0063] The Z score is used to check whether there are outliers in the data in the historical landslide training dataset, and the outliers are processed using scaling and normalization methods.
[0064] In some embodiments, the historical landslide training data set is divided into three parts: training set, validation set and test set according to 8:1:1. The training set is mainly used to train the model, and the training error is minimized by adjusting the model parameters and learning algorithm. The validation set is used to evaluate the performance of the model during the model training process, assist in selecting the best model parameters and configuration, and prevent the model from overfitting on the training set. The test set is used to evaluate the performance of the final model to verify the generalization ability and effect of the model in practical applications.
[0065] Specifically, the optimization training of the initial landslide warning model also includes: based on the specific prediction value y′ i and the true value y i The evaluation module is input to determine the evaluation indicators, which include MAE, MSE and RMSE, and the calculation methods are as follows:
[0066]
[0067]
[0068]
[0069] Among them, n represents the total number of samples, y i represents the true value, y i ' indicates a specific predicted value.
[0070] Figure 4 This is a comparison chart of evaluation parameters of the landslide disaster warning method based on the stacked autoencoder described in the embodiment of the present invention and the original autoencoder on the test set, which can intuitively show the effectiveness and advantages of the method of the present invention in landslide disaster warning.
[0071] In some embodiments, a neural network architecture, loss function, and hyperparameters for landslide disaster warning are established. During the training process, the landslide warning model can be configured in the following manner:
[0072] Model training includes:
[0073] Set the hyperparameters input_size: 10, hidden_size: 64, learning rate: 0.001, output_size: 1, number of epochs: 50, loss: 'mean_squared_error', psequence_length: 5, batch_size: 1, num_rules: 4, meta_hidden_size: 10, optimizer: Adam
[0074] Set training parameters: The initial learning rate is set to 0.001.
[0075] Model preservation: MAE, MSE and RMSE are used as evaluation indicators. During network training, a test set test is performed every time the weight is updated. The best model is finally saved through MAE, MSE and RMSE in the test set. Thus, a trained landslide warning model is obtained.
[0076] Using stacked autoencoders to predict and warn landslide displacement;
[0077] The encoder continuously optimizes its feature extraction capability by stacking meta-models to learn the deep features of the data, further refining the feature space transmitted from the upper layer, so that the autoencoder can better capture the deep data structure and patterns.
[0078] Figure 5 A loss graph for each round of training of the landslide disaster early warning method based on the original autoencoder according to an embodiment of the present invention; Figure 6 This is a loss graph for each round of training of the landslide disaster early warning method based on stacked autoencoders described in an embodiment of the present invention. Figure 5 and Figure 6 In comparison, the stacked autoencoder-based method shows a faster convergence speed, indicating that it can learn the characteristics of the data more effectively during the training process, thereby proving its superiority and effectiveness in the landslide disaster warning task.
[0079] This embodiment establishes a landslide disaster warning method based on stacked autoencoders, which can achieve accurate prediction and warning of mine landslide disasters. The autoencoder is suitable for landslide disaster warning tasks, can automatically extract key information from a large amount of data through its powerful feature encoding ability, and has the ability to effectively process and identify abnormal data patterns.
[0080] In this embodiment, model prediction and early warning include: data processing: first, the landslide monitoring data is obtained, and then the abnormal values in the data are processed; stacking metamodel: the present invention stacks the metamodel layer in the encoder FNN. The metamodel layer can self-regulate and optimize the model by learning deeper feature representations from the output of the FNN. The FNN first processes the data based on fuzzy logic and uses a Gaussian function to calculate the membership of the input features. Subsequently, the FNN converts the output of the fuzzy layer into a feature space that is easy for the model to understand. Finally, the metamodel acts as a decision layer to further process the feature space and optimize the received data features. Perform prediction and early warning: output the prediction results of the landslide displacement to complete the model prediction and early warning.
[0081] In this embodiment, a stacked autoencoder is used to predict the landslide displacement, so that it has the ability of self-regulation and optimization, reduces the model deviation, improves the final output of the model, and uses the quartile method to determine the threshold to complete the early warning; the MAE, MSE and RMSE obtained after each training are output to observe whether the relevant indicators meet the requirements, and finally save a group of models with the best training effect.
[0082] The autoencoder is optimized by stacking metamodel layers in the encoder FNN, and the acquired data is input into the trained stacked autoencoder for landslide displacement prediction and disaster warning.
[0083] Figure 3 It is an exemplary flow chart of a landslide hazard warning system based on stacked autoencoders according to some embodiments of this specification.
[0084] In some embodiments, the present invention discloses a landslide disaster warning system based on a stacked autoencoder, characterized in that the landslide disaster warning system includes the following modules:
[0085] An acquisition module is used to acquire landslide monitoring data of a target mountain;
[0086] A preprocessing module, used for preprocessing the landslide monitoring data to obtain landslide data characteristics and landslide displacement thresholds;
[0087] A prediction module, used for inputting the landslide data characteristics and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained based on the optimization training of the initial landslide early warning model, and is composed of a fuzzy layer based on fuzzy logic, a linear layer for processing linear transformation, and a decision layer constructed based on a stacked metamodel;
[0088] The determination module is used to determine the landslide warning result based on the predicted value of the landslide displacement.
[0089] In some embodiments, the present invention also includes a landslide disaster warning device based on a stacked autoencoder, including a processor, wherein the processor is used to execute a landslide disaster warning method based on a stacked autoencoder.
[0090] In some embodiments, the present invention also includes a computer-readable storage medium, which stores computer instructions. When the computer reads the computer instructions in the storage medium, the computer executes the landslide hazard warning method based on the stacked autoencoder.
[0091] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.
[0092] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.
[0093] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.
[0094] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.
[0095] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values is as accurate as possible within the feasible range.
[0096] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.
[0097] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.
Claims
1. A landslide disaster early warning method based on stacked autoencoders, It is characterized in that The following steps are involved: S1: Obtain landslide monitoring data of the target mountain; S2: preprocessing the landslide monitoring data to obtain landslide characteristic data and landslide displacement threshold; S3: inputting the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained based on the optimization training of the initial landslide early warning model, and is composed of a fuzzy layer based on fuzzy logic, a linear layer for processing linear transformation, and a decision layer constructed based on a stacked metamodel; S4: Determine a landslide warning result based on the predicted value of the landslide displacement.
2. The landslide disaster early warning method based on stacked autoencoders according to claim 1, It is characterized in that The landslide monitoring data described in S1 include: Through landslide monitoring sensors, the displacement of the landslide body, rainfall, effective horizontal stress, vertical displacement of the monitoring point, and geological and environmental data of the target mountain are obtained to construct a landslide dataset.
3. The landslide disaster early warning method based on stacked autoencoders according to claim 2, It is characterized in that The pre-processing in S2 specifically comprises the following steps: S21: determining outliers in the landslide data set based on a Z-score method; S22: Determine the landslide displacement threshold by quartile method.
4. The landslide disaster early warning method based on stacked autoencoders according to claim 3, It is characterized in that The S21 determines the outliers in the landslide data set based on the Z-score method, and further includes: For each observation in the landslide dataset, calculate its Z score; Determine whether the observed value is an outlier based on the absolute value of the Z score; According to the distribution and characteristics of all observation data in the data set, the Z-score normalization method is selected to deal with outliers.
5. The landslide disaster early warning method based on stacked autoencoders according to claim 3, It is characterized in that The S22 determines the landslide displacement threshold by using the quartile method, and further includes: For each observation value in the landslide data set, the lower quartile Q1 and the upper quartile Q3 of the observation value are calculated; Data points outside the range [Q1-1.5*IQR,Q3+1.5*IQR] are considered outliers.
6. The landslide disaster early warning method based on stacked autoencoders according to claim 5, It is characterized in that The step S3 inputs the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, which specifically includes the following steps: S31: Input the preprocessed landslide feature data into the fuzzy layer, and use the Gaussian function to calculate the membership of the input feature: Among them, R i is the fuzzy rule parameter vector, x is the input data vector, γ(x) i is the membership degree of the input data x to the i-th rule; S32: Input the membership of the output of the fuzzy layer into the linear layer for transformation: f(x)=ω·γ(x)+b, Among them, ω is the weight matrix, γ(x) i is the membership vector output by the fuzzy layer, and f(x) is the feature space output after linear transformation; S33: Use the ReLU activation function to increase nonlinearity: ReLU(f(x)) i =max(0,f(x) i ), Among them, f(x) i is the i-th element of the linear layer output vector, and the ReLU function sets all negative values to 0; S34: Input the output of the linear layer into the decision layer constructed based on the stacked meta-model to further process the feature space: M=W i ·ReLU(f(x)) i +B i , Among them, W i is the weight matrix of the meta-model layer, B i is the bias vector of the meta-model layer and M is the predicted value of the final output of the encoder.
7. The landslide disaster early warning method based on stacked autoencoders according to claim 1, It is characterized in that The optimization training process of the landslide early warning model includes: obtaining a historical landslide data set, wherein the historical landslide data set includes multiple groups of historical landslide displacement characteristic data; Inputting the historical landslide data set into an initial landslide early warning model to obtain a training prediction value, wherein the initial landslide early warning model is an autoencoder constructed based on a stacking metamodel; Based on the training prediction value and the true value input into the evaluation module, the evaluation index is determined, and the evaluation index includes MAE, MSE and RMSE.
8. A landslide disaster early warning system based on stacked autoencoders, It is characterized in that The landslide disaster early warning system includes the following modules: An acquisition module is used to acquire landslide monitoring data of a target mountain; A preprocessing module, used for preprocessing the landslide monitoring data to obtain landslide characteristic data and landslide displacement threshold; A prediction module, used for inputting the landslide characteristic data and the landslide displacement threshold into a landslide early warning model to obtain a predicted value of the landslide displacement, wherein the landslide early warning model is obtained by optimizing and training a fuzzy layer, a linear layer, and a decision layer constructed based on a stacked metamodel; The determination module is used to determine the landslide warning result based on the predicted value of the landslide displacement.
9. A landslide disaster early warning device based on a stacked autoencoder, comprising a processor, wherein the processor is used to execute the landslide disaster early warning method based on a stacked autoencoder according to any one of claims 1 to 7.
10. A computer-readable storage medium storing computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the landslide disaster early warning method based on stacked autoencoders according to any one of claims 1 to 7.
