A landslide displacement prediction method using Markov attention mechanism

Abstract

The present invention discloses a landslide displacement prediction method using a Markov attention mechanism, which relates to the field of landslide displacement feature extraction and high-precision prediction. The method obtains the landslide displacement time series of the research area and inputs the landslide displacement time series into the landslide displacement prediction model; in the landslide displacement prediction model, a landslide displacement Markov matrix is constructed based on the landslide displacement time series; the later elements in the landslide displacement time series appear more times in the landslide displacement Markov matrix; the landslide displacement Markov matrix is subjected to feature extraction through the Markov matrix information extraction model to obtain the landslide displacement change features; the landslide displacement time series and the landslide displacement change features are added to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series; the output sequence is input into a non-linear model to obtain the landslide displacement prediction value of the research area. This method can improve the prediction accuracy of landslide displacement.

Description

Technical Field

[0001] The present invention relates to the field of landslide displacement feature extraction and high-precision prediction, and particularly relates to a landslide displacement prediction using a Markov attention mechanism. Background Art

[0002] Landslide displacement prediction is a crucial part in landslide disaster analysis. It mainly builds a regression model between influencing factor data or historical landslide displacement data and future data based on the regression idea in time series analysis, and then uses optimization methods such as least squares and gradient descent to determine the model parameters, thereby obtaining a landslide displacement prediction model. The landslide process is relatively complex. Considering other influencing factors such as rainfall and temperature and establishing a landslide displacement prediction model by combining external influencing factors is called multivariate landslide displacement prediction. However, different landslide scenarios, internal compositions, and environments will all result in poor generalization performance of the multivariate model in different landslide displacement prediction scenarios.

[0003] Currently, landslide displacement prediction methods can be mainly divided into two categories: traditional prediction methods and data-driven methods. Traditional prediction methods mainly use algorithms such as grey prediction models that require less data volume and have higher ultra-short-term prediction accuracy. Therefore, data-driven models are the most commonly used models in landslide displacement prediction. Commonly used data-driven models are based fitting models, traditional machine learning models, and deep learning models. Currently, the most widely used in landslide displacement prediction is the prediction method based on deep learning. Such methods transform the feature engineering with high requirements into automated neuron learning, improving the generalization performance and accuracy of model prediction.

[0004] However, existing landslide displacement prediction methods mainly optimize the model itself and do not analyze and model the characteristics of landslide displacement, resulting in low prediction accuracy of landslide displacement. Summary of the Invention

[0005] Based on this, it is necessary to provide a landslide displacement prediction method using a Markov attention mechanism for the above technical problems. This method can improve the prediction accuracy of landslide displacement.

[0006] The present invention adopts the following technical solutions:

[0007] The present invention provides a landslide displacement prediction method using a Markov attention mechanism, including:

[0008] Obtain the landslide displacement time series of the research area and input the landslide displacement time series into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model;

[0009] In the landslide displacement prediction model, a landslide displacement Markov matrix is constructed based on the landslide displacement time series; the later elements in the landslide displacement time series appear more frequently in the landslide displacement Markov matrix;

[0010] Feature extraction is performed on the landslide displacement Markov matrix through the Markov matrix information extraction model to obtain the landslide displacement change characteristics; the Markov matrix information extraction model is constructed by multiple cascaded Markov attention mechanisms;

[0011] The landslide displacement time series and the landslide displacement change characteristics are added together to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series;

[0012] The output sequence is input into the nonlinear model to obtain the landslide displacement prediction value of the research area.

[0013] Preferably, the training process of the landslide displacement prediction model includes:

[0014] Obtain the historical landslide displacement time series and input the historical landslide displacement time series into the initial landslide displacement prediction model;

[0015] In the initial landslide displacement prediction model, a historical landslide displacement Markov matrix is constructed based on the historical landslide displacement time series;

[0016] The historical landslide displacement Markov matrix is extracted and processed through multiple cascaded Markov attention mechanisms to obtain the historical extraction result;

[0017] The historical landslide displacement time series and the historical extraction result are added together to obtain the historical output sequence after the Markov attention mechanism processes the historical landslide displacement time series;

[0018] The historical output sequence is input into the nonlinear model to obtain the predicted value of the initial landslide displacement prediction model for the historical landslide displacement time series;

[0019] Obtain the observed value of the historical output sequence;

[0020] Calculate the mean square error between the predicted value and the observed value, and adjust the parameters of the initial landslide displacement prediction model until the error is minimized to obtain the landslide displacement prediction model.

[0021] Preferably, the landslide displacement time series is:

[0022] X = {x1, x2, x3, …, x n};

[0023] where X is the landslide displacement time series, and x i represents the cumulative landslide displacement at the i-th moment, and i takes values from 1 to n;

[0024] The Markov matrix of landslide displacement is as follows:

[0025]

[0026] Among them, MarkovMatrix X is the Markov matrix of landslide displacement, and x i represents the cumulative landslide displacement at the i-th moment, and the value of i ranges from 1 to n.

[0027] Preferably, the Markov attention mechanism includes one-dimensional temporal convolution, one-dimensional pooling, and an activation function. The implementation process of the Markov attention mechanism specifically includes:

[0028] Input the input information of the Markov attention mechanism into the one-dimensional temporal convolution module to obtain the output of the one-dimensional temporal convolution module;

[0029] Input the output of the one-dimensional temporal convolution module into the one-dimensional pooling module to obtain the pooling result;

[0030] Input the pooling result into the activation function to obtain the output result of the Markov attention mechanism.

[0031] Preferably, the non-linear model is:

[0032] NLinear:R M →R N ;

[0033] Among them, M represents the length of the input sequence of the non-linear model, and N represents the length of the prediction sequence of the non-linear model.

[0034] The present invention provides a landslide displacement prediction device using a Markov attention mechanism, including:

[0035] An acquisition module for acquiring the time series of landslide displacement in the research area and inputting the time series of landslide displacement into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model;

[0036] A construction module for constructing a landslide displacement Markov matrix based on the time series of landslide displacement in the landslide displacement prediction model; the elements in the time series of landslide displacement that are later in time appear more times in the landslide displacement Markov matrix;

[0037] An extraction module for extracting the characteristics of the landslide displacement change from the landslide displacement Markov matrix through the Markov matrix information extraction model; the Markov matrix information extraction model is constructed by multiple cascaded Markov attention mechanisms;

[0038] The first determination module is configured to add the landslide displacement time series and the landslide displacement change characteristics to obtain an output series after the Markov attention mechanism processes the landslide displacement time series.

[0039] The second determination module is configured to input the output series into a non-linear model to obtain the landslide displacement prediction value of the research area.

[0040] The present invention provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the above-mentioned landslide displacement prediction method using a Markov attention mechanism.

[0041] The present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned landslide displacement prediction method using a Markov attention mechanism.

[0042] The above at least one technical solution adopted by the present invention can achieve the following beneficial effects:

[0043] First, by constructing a Markov matrix based on the landslide displacement time series, the information contained is gradually reduced from back to front in time, ensuring that the data closer to the prediction point contributes more information to the landslide displacement prediction model, greatly improving the ability of the landslide displacement prediction model to extract information near the prediction point. Second, multiple Markov attention mechanisms are used to extract and process the Markov matrix based on the landslide displacement time series, and the result of the extraction and processing is added to the landslide displacement time series to obtain the output of the Markov attention mechanism, combining single feature learning and inference prediction, and establishing a landslide displacement prediction model combining a Markov attention mechanism and a non-linear model. This method can improve the prediction accuracy of landslide displacement. Description of the Drawings

[0044] The drawings described herein are used to provide a further understanding of the present invention, and constitute a part of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0045] Figure 1 It is a schematic flow chart of a landslide displacement prediction method using a Markov attention mechanism provided by the present invention;

[0046] Figure 2 It is a working flow chart of the Markov attention mechanism provided by the present invention;

[0047] Figure 3 It is a working flow chart of a specific embodiment provided by the present invention;

[0048] Figure 4 Schematic diagram of the data after preprocessing the landslide data at point HF05 in the specific embodiment of the present invention;

[0049] Figure 5 Schematic diagram of the data after preprocessing the landslide data at point HF08 in the specific embodiment of the present invention;

[0050] Figure 6 Schematic diagram of the data after preprocessing the landslide data at point HF09 in the specific embodiment of the present invention;

[0051] Figure 7 Comparison diagram of the prediction performance of the MANLinear model provided by the present invention in the displacement mutation stage;

[0052] Figure 8 Prediction residual diagram of different landslide displacement prediction models provided by the present invention in the case of landslide displacement prediction mutation;

[0053] Figure 9 Comparison diagram of the improvement rate of the neural network accuracy of the MANLinear model provided by the present invention under prediction tasks with different data volumes;

[0054] Figure 10 Comparison diagram of the weight results at different positions of the neural network after training provided by the present invention;

[0055] Figure 11 Result diagram of the data sensitivity analysis of the MANLinear model provided by the present invention;

[0056] Figure 12 Schematic diagram of a landslide displacement prediction device using a Markov attention mechanism provided by the present invention;

[0057] Figure 13 Schematic diagram of a computer device for implementing a landslide displacement prediction method using a Markov attention mechanism provided by the present invention. Detailed implementation manners

[0058] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0059] Devices such as desktop computers, servers, and laptop computers that can execute the solution of the present invention. For the sake of convenience of description, only the server is used as the execution subject for description below.

[0060] The technical solutions provided by the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0061] Figure 1 The following is a schematic flow chart of a landslide displacement prediction method using the Markov attention mechanism in the present invention, which specifically includes the following steps:

[0062] S101: Obtain the landslide displacement time series of the research area, and input the landslide displacement time series into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model.

[0063] In an exemplary embodiment, the landslide displacement time series of the research area is shown in Formula (1):

[0064] X = {x1, x2, x3, …, x n} (1);

[0065] where X is the landslide displacement time series, and x i represents the cumulative landslide displacement at the i-th moment, and i takes values from 1 to n.

[0066] Specifically, the landslide displacement has very obvious autocorrelation characteristics, and there will be different landslide inducing factors under different geological and environmental conditions. Therefore, the research on the autocorrelation of landslide displacement is more fundamental and has better universality.

[0067] In an exemplary embodiment, the training process of the landslide displacement prediction model includes: obtaining the historical landslide displacement time series, and inputting the historical landslide displacement time series into the initial landslide displacement prediction model; in the initial landslide displacement prediction model, constructing a historical landslide displacement Markov matrix based on the historical landslide displacement time series; performing extraction processing on the historical landslide displacement Markov matrix through a series of multiple Markov attention mechanisms to obtain a historical extraction result; adding the historical landslide displacement time series and the historical extraction result to obtain a historical output sequence after the Markov attention mechanism processes the historical landslide displacement time series; inputting the historical output sequence into the non-linear model to obtain the predicted value of the initial landslide displacement prediction model for the historical landslide displacement time series; obtaining the observed value of the historical output sequence; calculating the mean square error between the predicted value and the observed value, and adjusting the parameters of the initial landslide displacement prediction model until the error is minimized to obtain the landslide displacement prediction model.

[0068] The calculation method of the mean square error between the predicted value and the observed value is shown in Formula (2):

[0069]

[0070] where Y is the observed value of the landslide displacement, is the predicted value of the landslide displacement model, is the mean square error between the predicted value and the observed value of landslide displacement, N is the number of observed values of landslide displacement, and Y i respectively represent the predicted value of the i-th landslide displacement model and the observed value of the i-th landslide displacement.

[0071] Specifically, the historical landslide displacement time series is used as the training set; the initial landslide displacement prediction model is trained through the training set, and the mean square error between the predicted value of the initial landslide displacement prediction model and the observed value of the historical output sequence is calculated; by adjusting the parameters of the initial landslide displacement prediction model until the mean square error is minimized, a landslide displacement prediction model is obtained.

[0072] Specifically, in the embodiment of the present invention, the error calculation formula between the predicted value and the true value of the landslide displacement prediction model for the historical landslide displacement sequence is formula (2). The mean square error is used to evaluate the prediction accuracy of the landslide displacement prediction model, and then the Adam optimization algorithm including learning rate adaptation and momentum gradient descent is used to adjust the model parameters, the convergence of the model error is completed on the training set, and the generalization error of the model is tested on the test set. Once the landslide displacement prediction model is trained, the model parameters can be fixed and deployed in the actual landslide displacement prediction.

[0073] S102: In the landslide displacement prediction model, a landslide displacement Markov matrix is constructed based on the landslide displacement time series; the elements in the landslide displacement time series that are later in time appear more times in the landslide displacement Markov matrix.

[0074] In an exemplary embodiment, the landslide displacement Markov matrix is shown in formula (3):

[0075]

[0076] where MarkovMatrix X is the landslide displacement Markov matrix, x i represents the cumulative landslide displacement at the i-th moment, and the value of i ranges from 1 to n.

[0077] Specifically, the data with earlier time nodes in the landslide displacement time series appear fewer times in the landslide displacement Markov matrix, and the data with later time nodes appear more times in the landslide displacement Markov matrix and provide more information.

[0078] S103: The landslide displacement Markov matrix is feature-extracted by the Markov matrix information extraction model to obtain the landslide displacement change feature; the Markov matrix information extraction model is constructed by multiple cascaded Markov attention mechanisms.

[0079] In an exemplary embodiment, the Markov attention mechanism includes one-dimensional temporal convolution, one-dimensional pooling, and an activation function. The implementation process of the Markov attention mechanism specifically includes: inputting the input information of the Markov attention mechanism into the one-dimensional temporal convolution module to obtain the output of the one-dimensional temporal convolution module; inputting the output of the one-dimensional temporal convolution module into the one-dimensional pooling module to obtain the pooling result; and inputting the pooling result into the activation function to obtain the output result of the Markov attention mechanism.

[0080] Specifically, the input of the first Markov attention mechanism in the Markov matrix information extraction model is the landslide displacement Markov matrix. Subsequently, the Markov matrix is processed through multiple cascaded Markov attention mechanisms, and the output of the previous Markov attention mechanism is the input of the next Markov attention mechanism.

[0081] The one-dimensional temporal convolution is shown in Equation (4):

[0082]

[0083] where x is the given input sequence, w represents the convolution kernel, K is the length of the convolution kernel, y is the output of the one-dimensional temporal convolution, y is denoted as Conv1d(), i is the convolution kernel index, and t is the time point of the output sequence.

[0084] Pooling is an important means to reduce the data dimension and prevent overfitting, and it mainly includes mean pooling and max pooling.

[0085] Max pooling is shown in Equation (5):

[0086] y j =max(x j ,x j+1 ,…,x j+K-1 ) (5);

[0087] where y j is the output value after the pooling operation, that is, the maximum value of the pooling window, x j is the starting value of the pooling window, j is the starting position of the pooling window, and K is the length of the pooling window.

[0088] Mean pooling is shown in Equation (6):

[0089]

[0090] where x is the given input time series, K is the length of the pooling window, y is the pooling result, and for convenience of representation, the pooling result is denoted as Pooling1d(x).

[0091] The activation function is the key for the neural network to perform non - linear mapping. The commonly used ReLU function is shown in Formula (7):

[0092] ReLU(x) = max(0, x) (7);

[0093] Where ReLU is the activation function and x is the given input time series.

[0094] Then the Markov attention mechanism is shown in Formula (8) and Formula (9):

[0095] G = ReLU(CnnBlock(CnnBlock(MarkovMatrix X )) (8);

[0096] CnnBlock(x) = Pooling1d(ReLU(Conv1d(x))) (9);

[0097] Where G is the extraction result, ReLU is the activation function, MarkovMatrix X is the landslide displacement Markov matrix, Pooling1d is the pooling result, and Conv1d is the one - dimensional time - series convolution operation.

[0098] S104: Add the landslide displacement time series and the landslide displacement change characteristics to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series.

[0099] Specifically, in Formula (8), G is the extraction result. Add it to the landslide displacement time series to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series, as shown in Formula (10):

[0100]

[0101] Where is the output sequence, G is the extraction result, and X is the landslide displacement time series.

[0102] Specifically, the overall calculation process of the Markov attention mechanism is as Figure 2 shown. The landslide displacement input sequence is the landslide displacement time series mentioned in the present invention. Based on the landslide displacement input sequence, a Markov matrix is constructed. Perform multi - layer one - dimensional convolution, activation function ReLU, and one - dimensional pooling operations on the constructed Markov matrix to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series.

[0103] S105: Input the output sequence into a non - linear model to obtain the landslide displacement prediction value of the research area.

[0104] In an exemplary embodiment, the non - linear model is shown in Formula (11):

[0105] NLinear:R M →R N (11);

[0106] Where M is the length of the input sequence of the non - linear model, and N is the length of the prediction sequence of the non - linear model.

[0107] Specifically, the non - linear model is a mapping from the input data dimension to the output data dimension. In the present invention, the input sequence of the non - linear model is the output sequence of the Markov attention mechanism, and the output sequence of the non - linear model is the predicted value of the landslide displacement time series.

[0108] In an exemplary embodiment, the predicted value of the landslide displacement prediction model for the landslide displacement time series is shown in Formula (12):

[0109]

[0110] Where is the predicted value of the trained landslide displacement model for the landslide displacement time series, is the output sequence of the Markov attention mechanism of the landslide displacement time series, and NLinear is the non - linear model.

[0111] Specifically, the output sequence of the Markov attention mechanism is predicted through the non - linear model to obtain the predicted value of the landslide displacement model for the landslide displacement time series.

[0112] The present invention innovatively models the Markov characteristics of the landslide displacement change process, enabling the landslide displacement prediction model to more directly focus on the information of adjacent points of the prediction point in the input information. Combining with the subsequent prediction model realizes a higher - precision landslide displacement prediction. The present invention designs a Markov attention mechanism to model the Markov process, combines it with the existing neural network prediction model, greatly improves the prediction accuracy of the landslide displacement model, and significantly improves the accuracy and stability of the prediction in the rapid - change stage of the landslide displacement.

[0113] The present invention greatly improves the prediction accuracy of the landslide displacement by considering the Markov characteristics of the landslide displacement change process. According to the high - precision landslide displacement prediction value, further combining with work such as landslide threshold discrimination, etc., it can be actually deployed in dangerous slopes or potential hazard slopes, providing a scientific basis for economic development, urban and rural construction, land use planning, geological disaster prevention and control, etc.

[0114] In an exemplary embodiment, as Figure 3The specific working flowchart of the solution of the present invention is provided as follows. First, cumulative time-series monitoring displacement data of the landslide is obtained, and the displacement data is preprocessed to obtain the preprocessed data. The preprocessing specifically includes gross error rejection, trend extraction, and data resampling. Since there are still a large number of significant gross errors in the original data due to the acquisition method, gross error rejection is required. At the same time, due to the limitations of the positioning method itself, its results have obvious fluctuations, so trend extraction is needed. The sampling interval of the original data is 5 minutes. This dense sampling interval leads to significant data redundancy, and at the same time, this time interval cannot meet the requirements of landslide displacement early warning and forecasting, so data resampling is needed. The preprocessed data is segmented to obtain a test set, a validation set, and a training set. Then, according to the training set, a landslide displacement prediction model using the Markov attention mechanism (Markov AttentionNLinear, MANLinear) proposed by the present invention is trained, and the MANLinear model is optimized using the validation set to obtain a trained MANLinear model. Finally, the test set is input into the trained MANLinear model to obtain the prediction result of the MANLinear model for the test set.

[0115] In another exemplary embodiment, the experimental area selected by the present invention is a typical loess landslide area. In order to better understand the movement and deformation conditions of the typical landslide body, a high-precision Beidou and Global Navigation Satellite System (GNSS) monitoring network is arranged in this area to obtain long-time series deformation monitoring information of the typical landslide body. Landslide data of three collapse events are selected: the landslide at point HF08 on March 26, 2019, the landslide at point HF09 on January 29, 2021, and the landslide data at point HF05 on July 8, 2021. Given the complexity of the research area environment and the problem of gross errors in the monitored time-series displacement, the obtained landslide displacement monitoring data contains significant gross errors. At the same time, in order to verify the test results of the model at different time scales, data preprocessing work such as gross error rejection, trend extraction, and resampling is carried out on the original data, such as Figure 4 Data after preprocessing of the landslide data at point HF05, such as Figure 5 Data after preprocessing of the landslide data at point HF08, such as Figure 6 Data after preprocessing of the landslide data at point HF09.

[0116] In this embodiment, the splitting ratio of the training set, validation set, and test set is 8:1:1. To ensure sufficient information input and consider the model complexity at the same time, 7 historical landslide displacement data are selected to predict 1 future landslide displacement data. Table 1 shows the number of training sets of different landslide monitoring points at different time intervals after organizing the datasets of HF05, HF08, and HF09. The prediction performance of different models is evaluated by calculating the root mean square error (RMSE) between the prediction results and the true values in the test set.

[0117] Table 1

[0118]

[0119] To verify the superiority of the proposed MANLinear in this invention and compare the prediction accuracies of different models, in addition to MANLinear and the basic model (NLinear), time series prediction models in the field of deep learning including Non-stationary Transformer, FEDformer, Autoformer, and Informer, as well as traditional machine learning algorithms including Support Vector Regression (SVR), Extreme Learning Machines (ELM), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) are selected. The prediction accuracies of different models are evaluated by calculating the RMSE of the above models in the test set.

[0120] In this embodiment, experiments are conducted on different time resolutions of the landslide displacement monitoring points of HF05, HF08, and HF09. The RMSE errors of different models on the test set are shown in Table 2.

[0121] Table 2

[0122]

[0123] For the HF05 landslide monitoring point with relatively sufficient data volume, the RMSE of the MANLinear model is better than all other comparison models. Compared with the basic model NLinear before improvement, the accuracy is improved by 91.96%, 74.59%, and 63.68% at time intervals of 6 hours, 12 hours, and 24 hours respectively; for the HF08 data set with insufficient data volume, the RMSE errors of the MANLinear model are also lower than all other comparison models. Compared with the basic model NLinear, the accuracy is improved by 91.73%, 82.08%, and 78.75% at time intervals of 6 hours, 12 hours, and 24 hours respectively; for the prediction of the HF09 landslide monitoring point, in the case of relatively scarce training data, MANLinear still has the smallest RMSE among all model comparisons. Compared with the basic model NLinear, the accuracy is improved by 64.24%, 53.82%, and 42.92% at time intervals of 6 hours, 12 hours, and 24 hours respectively.

[0124] Figure 7 To visually display the prediction performance of the MANLinear model in the displacement mutation stage, the test results of the 6-hour time interval for HF05, HF08, and HF09 are visualized. To more clearly display the comparison results, in addition to the model proposed in the present invention, only the test results of the two models with the second-best prediction accuracy in Table 2 are selected for comparison. For the HF05, HF08, and HF09 points, the other two models with the second-best prediction accuracy are LSTM and GRU, LSTM and ELM, LSTM and ELM.

[0125] Figure 8 The prediction residuals of different landslide displacement prediction models in the case of landslide displacement prediction mutation are given. Further analyze the different prediction performances of different models in the landslide displacement mutation with more data volume. Figure 8 In Figure (a), there is a mutation in the middle of HF05. Figure 8 In Figure (b), there is a mutation at the end of HF05. Figure 8 In Figure (c), there is a mutation at the end of HF08. Calculate the absolute value of the prediction residual of each model. Figure 8 It shows that the MANLinear model is significantly better than all comparison models in the prediction of landslide displacement mutation. Its average residual and residual range are significantly lower than the selected comparison models, about 1 / 2 to 1 / 4 of the comparison models, indicating that the model proposed in the present invention has more excellent prediction accuracy and stability in the case of landslide displacement mutation. At the same time, this result also verifies the correctness of the design of the Markov attention mechanism.

[0126] Figure 9 A comparison chart of the accuracy improvement rate of the neural network for the MANLinear model proposed in the present invention under prediction tasks with different data volumes is given.Figure 9 It shows that for the HF08 landslide monitoring point, the improvement rate of the model prediction accuracy at different time intervals is significantly higher than that of the HF05 landslide monitoring point. Especially at the 24-hour time interval, the difference in the improvement rate of the model accuracy before and after improvement between HF08 and HF05 is about 20%. The above characteristics indicate that the Markov attention mechanism can still greatly improve the accuracy of the NLinear model when the data is relatively scarce; the improvement rate of the HF09 landslide monitoring point at different time intervals is significantly lower than that of HF08 and HF05, indicating that when the amount of data is too small, it will limit the prediction effect of the improved algorithm, but the MANLinear model can still achieve a relatively considerable improvement in accuracy.

[0127] Figure 10 The comparison chart of the weight results at different positions of the neural network is given. To verify the hypothesis in the above text that the Markov attention mechanism pays more attention to the information near the prediction point, a two-layer linear neural network without hidden layers is designed, and the 6-hour time interval of the HF05 data set with the largest amount of data is used as Figure 10 shown in Figure (a) and the 3-hour time interval as Figure 10 shown in Figure (b) for training. As Figure 10 shown, there are obvious differences in the weights at different positions. The weight of the position closer to the prediction point (the more backward) is larger, while the weight of the position farther from the prediction point (the more forward) is smaller, and the weight is close to 0. This result is consistent with the existing relevant research results, verifying the starting point of the design of the Markov attention mechanism.

[0128] Figure 11 The data sensitivity analysis results of the MANLinear model proposed in the present invention are given. Data-driven models tend to improve the model prediction accuracy with the increase in the amount of data, that is, as the training data increases, the model can learn the internal changes of the data more accurately. Therefore, for the same prediction task, the change of the error under different amounts of data reflects the data-driven performance of the model. Figure 11 Figure (a) in Figure 11 Figure (b) in Figure 11 Figure (c) in Figure 11It can be seen that as the amount of data increases, its accuracy does not decrease steadily. Instead, there is a situation where it increases rather than decreases. Therefore, increasing the amount of data does not directly improve the accuracy of the basic model prediction. On the contrary, it may even make the model worse, indicating that the basic model is not an efficient data-driven model. Due to its simple structure, the basic model cannot fully extract the information contained in the data after increasing the amount of data, so it will instead reduce the prediction accuracy of the model. The improved model proposed in the present invention not only significantly improves the prediction accuracy of the basic model but also exhibits good data-driven characteristics, that is, its prediction accuracy steadily increases as the amount of data increases.

[0129] The MANLinear model proposed in the present invention first constructs a temporal Markov matrix to gradually reduce the information it contains from back to front, ensuring that the data closer to the prediction point contributes more information to the model. Then, it uses a stacked one-dimensional convolutional network to perform sliding convolution on the Markov matrix. Finally, the convolution result is added to a residual connection to obtain the output of the Markov attention mechanism. At the same time, by combining single representation learning and inference prediction, a landslide displacement prediction model combining the Markov attention mechanism and the NLinear model is established. The new model greatly improves the model's ability to extract information near the prediction point, and in the obvious Markov process such as landslide displacement, it greatly improves the prediction accuracy of landslide displacement and significantly improves the accuracy of landslide displacement prediction in the rapid change stage.

[0130] Traditional landslide displacement prediction models usually use the "historical landslide displacement time series" as the data for model learning. However, the historical landslide displacement time series does not include landslide displacement mutations, that is, landslide events. Therefore, the landslide displacement prediction model trained on historical data has a low prediction accuracy when encountering real landslide displacement mutations because deep learning models can only fit the features they have seen during the training stage. The Markov attention mechanism provided in the present invention can extract the information of the input sequence with emphasis. Since the landslide displacement process is a Markov process and there is a strong correlation between adjacent data, using the features extracted by the Markov attention mechanism for landslide displacement prediction can greatly improve the prediction accuracy in the landslide displacement mutation stage.

[0131] When applying the landslide displacement prediction method using the Markov attention mechanism provided by the present invention, it is not necessary to execute according to Figure 1 the order of the steps shown. The specific execution order of each step can be determined as needed, and the present invention does not limit this.

[0132] The above is a landslide displacement prediction method using the Markov attention mechanism provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding landslide displacement prediction device using the Markov attention mechanism, as shown in Figure 12 Figure [ID number not provided in the original, should be filled in according to the actual figure number].

[0133] Figure 12 Schematic diagram of a landslide displacement prediction device using a Markov attention mechanism provided by the present invention, including:

[0134] An acquisition module 1201, configured to acquire the landslide displacement time series of the research area and input the landslide displacement time series into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model.

[0135] A construction module 1202, configured to construct a landslide displacement Markov matrix based on the landslide displacement time series in the landslide displacement prediction model; the elements with later time in the landslide displacement time series appear more times in the landslide displacement Markov matrix.

[0136] An extraction module 1203, configured to extract features from the landslide displacement Markov matrix through the Markov matrix information extraction model to obtain landslide displacement change features; the Markov matrix information extraction model is constructed by a plurality of cascaded Markov attention mechanisms.

[0137] A first determination module 1204, configured to add the landslide displacement time series and the landslide displacement change features to obtain an output series after the Markov attention mechanism processes the landslide displacement time series.

[0138] A second determination module 1205, configured to input the output series into the non-linear model to obtain the landslide displacement prediction value of the research area.

[0139] For the specific limitations on a landslide displacement prediction device using a Markov attention mechanism, reference may be made to the limitations on a landslide displacement prediction method using a Markov attention mechanism in the above text, which will not be elaborated here. Each module in the above-mentioned landslide displacement prediction device using a Markov attention mechanism can be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned modules.

[0140] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program can be used to execute the above Figure 1 Provided landslide displacement prediction method using a Markov attention mechanism.

[0141] The present invention also provides Figure 13 Schematic diagram of the structure of the computer device shown in Figure 13As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include other hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the above Figure 1 A landslide displacement prediction method using a Markov attention mechanism provided.

[0142] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0143] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope recorded by the present invention.

Claims

1. A landslide displacement prediction method using a Markov attention mechanism, characterized in that, Including: Obtain the landslide displacement time series of the research area, and input the landslide displacement time series into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model; In the landslide displacement prediction model, construct a landslide displacement Markov matrix based on the landslide displacement time series; the later the time in the landslide displacement time series, the more times the element appears in the landslide displacement Markov matrix; Extract features from the landslide displacement Markov matrix through the Markov matrix information extraction model to obtain landslide displacement change features; The Markov matrix information extraction model is constructed by multiple cascaded Markov attention mechanisms; Add the landslide displacement time series and the landslide displacement change features to obtain the output sequence after the Markov attention mechanism processes the landslide displacement time series; Input the output sequence into the non-linear model to obtain the landslide displacement prediction value of the research area.

2. The method according to claim 1, characterized in that, The training process of the landslide displacement prediction model includes: Obtain the historical landslide displacement time series, and input the historical landslide displacement time series into the initial landslide displacement prediction model; In the initial landslide displacement prediction model, construct a historical landslide displacement Markov matrix based on the historical landslide displacement time series; Extract and process the historical landslide displacement Markov matrix through multiple cascaded Markov attention mechanisms to obtain a historical extraction result; Add the historical landslide displacement time series and the historical extraction result to obtain the historical output sequence after the Markov attention mechanism processes the historical landslide displacement time series; Input the historical output sequence into the non-linear model to obtain the prediction value of the initial landslide displacement prediction model for the historical landslide displacement time series; Obtain the observed value of the historical output sequence; Calculate the mean square error between the prediction value and the observed value, and adjust the parameters of the initial landslide displacement prediction model until the error is minimized to obtain the landslide displacement prediction model.

3. The method according to claim 1, wherein The landslide displacement time series is: ； Among them, X is the landslide displacement time series, represents the cumulative landslide displacement at the th moment, i takes values from 1 to n ; The landslide displacement Markov matrix is: ； Among them, is the Markov matrix of landslide displacement, represents the cumulative landslide displacement at the th moment, i takes values from 1 to n .

4. The method according to claim 1, wherein The Markov attention mechanism includes one-dimensional temporal convolution, one-dimensional pooling, and an activation function. The implementation process of the Markov attention mechanism specifically includes: Input the input information of the Markov attention mechanism into the one-dimensional temporal convolution module to obtain the output of the one-dimensional temporal convolution module; Input the output of the one-dimensional temporal convolution module into the one-dimensional pooling module to obtain the pooling result; Input the pooling result into the activation function to obtain the output result of the Markov attention mechanism.

5. A landslide displacement prediction device using a Markov attention mechanism, characterized in that, Including: An acquisition module for obtaining the landslide displacement time series of the research area and inputting the landslide displacement time series into the landslide displacement prediction model; the landslide displacement prediction model includes a Markov matrix information extraction model and a non-linear model; A construction module, which is used to construct a landslide displacement Markov matrix based on the landslide displacement time series in the landslide displacement prediction model; the later the time in the landslide displacement time series, the more times the element appears in the landslide displacement Markov matrix; An extraction module, which is used to extract features of the landslide displacement Markov matrix through a Markov matrix information extraction model to obtain landslide displacement change features; The Markov matrix information extraction model is constructed by a plurality of cascaded Markov attention mechanisms; A first determination module, which is used to add the landslide displacement time series and the landslide displacement change features to obtain an output series after the Markov attention mechanism processes the landslide displacement time series; A second determination module, which is used to input the output series into a non-linear model to obtain the landslide displacement prediction value of the research area.

6. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1 to 4 is implemented.

7. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method described in any one of claims 1 to 4 is implemented.

Images