A method and device for predicting seismic activity parameters

Abstract

This application provides a method and apparatus for predicting seismic activity parameters, comprising: acquiring seismic activity data and environmental data; preprocessing the seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data; fusing the preprocessed seismic activity data and environmental data to obtain fused features; inputting the fused features into a seismic activity parameter prediction model constructed based on a time-recurrent neural network, and having the seismic activity parameter prediction model output predicted seismic activity parameters. The model of this application fuses multi-source data affecting seismic activity for feature extraction, and improves the model based on the seismic activity parameter prediction task, thereby improving the accuracy of parameter prediction results.

Description

Technical Field

[0001] This application relates to the field of earthquake prediction technology, and in particular to a method and apparatus for predicting earthquake activity parameters. Background Technology

[0002] Seismic activity parameters are core indicators for describing the characteristics of seismic activity, revealing the laws of crustal movement, and assessing earthquake risk. Traditional methods for studying seismic activity parameters generally utilize historical earthquake data to construct statistical models, such as the Poisson model. These models assume that earthquakes occur as independent random events and calculate the probability of earthquakes of different magnitudes occurring within a certain future timeframe based on the frequency of historical earthquakes. For example, statistical analysis of earthquake data from the past 100 years in a certain region suggests that if an average of one earthquake of magnitude 5 or higher occurs annually in that region, the probability of an earthquake of magnitude 5 or higher occurring within the next 1, 5, and 10 years can be estimated using the Poisson model. However, seismic activity is not entirely independent; it is the result of the combined effects of multiple complex environmental factors, including crustal movement, underground fluids, and geophysical fields. Statistical models analyze a single type of data, failing to consider the influence of various environmental factors on seismic activity and unable to integrate multi-source data such as seismic activity data, crustal movement data, underground fluid data, and geophysical field data, resulting in low accuracy in predictions. Summary of the Invention

[0003] In view of this, the purpose of the embodiments of this application is to provide a method and apparatus for predicting seismic activity parameters.

[0004] To achieve the above objectives, embodiments of this application provide a method for predicting seismic activity parameters, including:

[0005] Acquire seismic activity data and environmental data; The seismic activity data and environmental data are preprocessed to obtain preprocessed seismic activity data and environmental data. The preprocessed seismic activity data and environmental data are fused to obtain fused features; The fused features are input into a seismic activity parameter prediction model constructed based on a time-recurrent neural network, and the seismic activity parameter prediction model outputs the predicted seismic activity parameters.

[0006] Optionally, the seismic activity parameter prediction model includes an input gate, a forget gate, and an output gate, wherein the input gate is represented as: (1) in, i t This represents the input gate output at the current time step t, where σ represents the activation function. w i This represents the weight matrix of the input gate.b i This represents the bias term of the input gate; h t-1 This indicates the hidden state of the previous time step. P t This represents the seismic activity parameter above a set threshold. ΔT t Anomaly is the time interval between the seismic event detected at the current time step t that is above a set threshold and the last seismic event that was above the set threshold. t This is the seismic anomaly index for the current time step t; The forget gate is represented as follows: (2) in, f t This represents the output of the forget gate at the current time step t. w f The weight matrix representing the forget gate, b f The bias term representing the forget gate; The output gate is represented as: (3) in, o t This indicates the output of the gate at the current time step t. w o This represents the weight matrix of the output gate. b o This represents the bias term of the output gate; The candidate memory units of the model are: (4) in, C t This represents the candidate memory cell at the current time step t. w c The weight matrix represents the candidate memory units. b c The bias term represents the candidate memory unit.

[0007] Optionally, the seismic activity parameter prediction model includes a Dropout layer. The Dropout rate of the Dropout layer is determined based on the activation value of the activation function of the previous layer. When the activation value is greater than the mean of the activation values, the Dropout rate is a first probability; when the activation value is less than the mean of the activation values, the Dropout rate is a second probability.

[0008] Optionally, the first probability is less than the second probability.

[0009] Optionally, the method further includes: Collect earthquake-related data and corresponding environmental data for the target area within a certain period of time; Earthquake-related data and corresponding environmental data are divided according to a preset time window to obtain the divided earthquake-related data and corresponding environmental data. Based on the segmented earthquake correlation data, earthquake activity parameters are calculated; The calculated seismic activity parameters and corresponding environmental data are preprocessed to obtain preprocessed seismic activity data and corresponding environmental data. Using the preprocessed seismic activity data and corresponding environmental data as data samples, the model constructed based on a time-recurrent neural network is trained to obtain a trained seismic activity parameter prediction model.

[0010] This application also provides a seismic activity parameter prediction device, including: The acquisition module is used to acquire seismic activity data and environmental data; The preprocessing module is used to preprocess the seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data. The fusion module is used to fuse the preprocessed seismic activity data and environmental data to obtain fused features; The prediction module is used to input the fused features into a seismic activity parameter prediction model constructed based on a time recurrent neural network, and the seismic activity parameter prediction model outputs the predicted seismic activity parameters.

[0011] Optionally, the seismic activity parameter prediction model includes an input gate, a forget gate, and an output gate, wherein the input gate is represented as: (1) in, i t This represents the input gate output at the current time step t, where σ represents the activation function. w i This represents the weight matrix of the input gate. b i This represents the bias term of the input gate; h t-1 This indicates the hidden state of the previous time step. P t This represents the seismic activity parameter above a set threshold. ΔT t Anomaly is the time interval between the seismic event detected at the current time step t that is above a set threshold and the last seismic event that was above the set threshold. tThis is the seismic anomaly index for the current time step t; The forget gate is represented as follows: (2) in, f t This represents the output of the forget gate at the current time step t. w f The weight matrix representing the forget gate, b f The bias term representing the forget gate; The output gate is represented as: (3) in, o t This indicates the output of the gate at the current time step t. w o This represents the weight matrix of the output gate. b o This represents the bias term of the output gate; The candidate memory units of the model are: (4) in, C t This represents the candidate memory cell at the current time step t. w c The weight matrix represents the candidate memory units. b c The bias term represents the candidate memory unit.

[0012] Optionally, the seismic activity parameter prediction model includes a Dropout layer. The Dropout rate of the Dropout layer is determined based on the activation value of the activation function of the previous layer. When the activation value is greater than the mean of the activation values, the Dropout rate is a first probability; when the activation value is less than the mean of the activation values, the Dropout rate is a second probability.

[0013] Optionally, the first probability is less than the second probability.

[0014] Optionally, the device further includes: The acquisition module is used to collect earthquake-related data and corresponding environmental data of the target area within a certain period of time. The segmentation module is used to segment earthquake-related data and corresponding environmental data according to a preset time window, so as to obtain the segmented earthquake-related data and corresponding environmental data. The calculation module is used to calculate seismic activity parameters based on the segmented seismic correlation data; The preprocessing module is used to preprocess the calculated seismic activity parameters and corresponding environmental data to obtain preprocessed seismic activity data and corresponding environmental data. The training module is used to train the model based on the time recurrent neural network using the preprocessed seismic activity data and corresponding environmental data as data samples, so as to obtain the trained seismic activity parameter prediction model.

[0015] As can be seen from the above description, the seismic activity parameter prediction method and apparatus provided in this application acquire seismic activity data and environmental data, preprocess the seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data, fuse the preprocessed seismic activity data and environmental data to obtain fused features, input the fused features into a seismic activity parameter prediction model constructed based on a time recurrent neural network, and output the predicted seismic activity parameters from the seismic activity parameter prediction model. The model in this application fuses multi-source data affecting seismic activity for feature extraction, improves upon the seismic activity parameter prediction task, enhances training efficiency, and increases the accuracy of prediction results. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the method flow of an embodiment of this application; Figure 2 This is a structural diagram of the device according to an embodiment of this application; Figure 3 This is a block diagram of the electronic device structure according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] like Figure 1 As shown in the embodiment of this application, a method for predicting seismic activity parameters is provided, including: S101: Acquire seismic activity data and environmental data; In this embodiment, seismic activity data and environmental data of the study area are acquired. The seismic activity data includes seismic activity parameters obtained from earthquake-related data, such as the maximum magnitude at a certain time scale, the α value reflecting the overall seismic activity level of the region, the b value representing the proportional relationship between the number of earthquakes of different magnitudes, seismic activity degree, seismic activity evolution index, average earthquake occurrence rate λ, and aftershock attenuation index p. The environmental data includes data that affect seismic activity, such as crustal movement data, underground fluid data, and geophysical field data.

[0021] S102: Preprocess the seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data; In this embodiment, the acquired seismic activity data and environmental data are preprocessed, including data cleaning, outlier removal or correction, format conversion, normalization and standardization, etc. The preprocessing removes noise and transforms the data into a form suitable for input into the model.

[0022] S103: The preprocessed seismic activity data and environmental data are fused to obtain fused features; In this embodiment, a feature fusion layer is used to fuse preprocessed seismic activity data and environmental data to obtain fused features. This feature fusion layer effectively integrates features from multiple sources and of multiple types, uncovering the inherent correlations between different features. Through weighted fusion, learnable dynamic weights are assigned to different features, enabling the model to automatically focus on features more important to the task. Feature fusion overcomes the limitations of single features, fully extracting the complementary knowledge contained in multi-source information, and providing more comprehensive and effective input for subsequent feature extraction.

[0023] S104: Input the fused features into the seismic activity parameter prediction model built on a time recurrent neural network, and output the predicted seismic activity parameters from the seismic activity parameter prediction model.

[0024] In this embodiment, after the multi-source data is fused into fused features by the feature fusion layer, the fused features are input into the seismic activity parameter prediction model. The seismic activity parameter prediction model extracts key features based on the fused features and predicts the seismic activity parameters for a certain period of time in the future based on the key features.

[0025] In some implementations, seismic activity parameter prediction models are built upon Long Short-Term Memory (LSTM) neural networks. LSTM layers perform temporal modeling of fused features, uncovering long-term dependencies that change over time. Through gating mechanisms such as forget gates, input gates, and output gates, as well as cell states, LSTM layers achieve selective memorization and forgetting of temporal information. The forget gate filters historical information, determining which past features need to be retained and which can be discarded, preserving information valuable for the current computation. The input gate fuses the features of the current time step with historical information, updates the cell state, and stores new, important information in memory. The output gate generates the hidden state for the current time step based on the current cell state and passes it to the next time step or subsequent layers.

[0026] In this embodiment, to adapt to the feature processing of seismic activity parameters, the input gate of the seismic activity parameter prediction model is represented as follows: (1) in, i t This represents the input gate output at the current time step t, where σ represents the activation function. w i This represents the weight matrix of the input gate. b i This represents the bias term of the input gate; h t-1 This indicates the hidden state of the previous time step. P tThis represents seismic activity parameters above a set threshold, such as a value of b that exceeds the set threshold. ΔT t Anomaly is the time interval between the seismic event detected at the current time step t that is above a set threshold and the last seismic event that was above the set threshold. t This is the seismic anomaly index at the current time step t.

[0027] The forget gate is represented as: (2) in, f t This represents the output of the forget gate at the current time step t. w f The weight matrix representing the forget gate, b f This represents the bias term for the forget gate.

[0028] The output gate is represented as: (3) in, o t This indicates the output of the gate at the current time step t. w o This represents the weight matrix of the output gate. b o This represents the bias term of the output gate.

[0029] The candidate memory units for the seismic activity parameter prediction model are: (4) in, C t This represents the candidate memory cell at the current time step t. w c The weight matrix represents the candidate memory units. b c The bias term represents the candidate memory unit.

[0030] In some embodiments, the seismic activity parameter prediction model uses multiple hidden units, the number of which determines the model's ability to capture complex dependencies. Simultaneously, a multi-layer LSTM stack approach can be employed to extract more abstract, higher-order temporal features through a deep structure. For scenarios requiring the retention of information at each time step, the LSTM layer outputs the hidden states of all time steps; if only the final result is needed, the hidden state of the last time step is output. Through this mechanism, the LSTM layer can effectively capture the relationships between data, providing the model with powerful modeling capabilities.

[0031] In some embodiments, the seismic activity parameter prediction model includes a Dropout layer. The Dropout rate of the Dropout layer is determined based on the activation value of the activation function of the previous layer. When the activation value is greater than the mean of the activation values, the Dropout rate is the first probability; when the activation value is less than the mean of the activation values, the Dropout rate is the second probability.

[0032] In this embodiment, the seismic activity parameter prediction model includes a fully connected layer used to perform nonlinear transformations and dimensional mapping on the temporal features output by the LSTM layer, compressing the high-dimensional temporal features into low-dimensional features closely related to the prediction task. The fully connected layer consists of multiple neurons, each connected to all neurons in the previous layer. A linear combination of features is achieved through a weight matrix, followed by a nonlinear transformation introduced by the ReLU activation function, enabling the model to learn the complex nonlinear relationships between features. The number of neurons in the fully connected layer typically decreases gradually from high to low. For example, the 64-dimensional features output by the LSTM layer are first mapped to 32 dimensions, and then further compressed to 16 dimensions, continuously refining key features while reducing feature dimensionality.

[0033] To prevent overfitting, a Dropout layer is introduced into the fully connected layer to randomly discard the outputs of some neurons, thereby enhancing the model's generalization ability. Based on the characteristics of seismic activity parameters, this application sets a dynamic Dropout rate based on the activation values ​​of these parameters. For neurons with activation values ​​higher than the mean, the dropout rate is set to a first probability (e.g., 0.3), while for neurons with activation values ​​lower than the mean, the dropout rate is set to a second probability (e.g., 0.6). This approach preserves key feature information while enhancing the network's generalization ability.

[0034] In some embodiments, the fully connected layer of the seismic activity parameter prediction model is connected to the output layer. The low-dimensional features output by the fully connected layer are mapped to interpretable task results through the output layer. The structural design of the output layer is closely related to the task type: in classification tasks, the number of neurons in the output layer is consistent with the number of categories, and the Softmax activation function is used to output the probability that a sample belongs to the positive class; in sequence prediction tasks, the output layer combines the temporal features output by the LSTM layer to output prediction results for multiple future time steps, and its dimension corresponds to the predicted time step size and the number of target features.

[0035] In some embodiments, the method further includes: Collect earthquake-related data and corresponding environmental data for the target area within a certain period of time; Earthquake-related data and corresponding environmental data are divided according to a preset time window to obtain the divided earthquake-related data and corresponding environmental data. Based on the segmented earthquake correlation data, earthquake activity parameters are calculated; The calculated seismic activity parameters and corresponding environmental data are preprocessed to obtain preprocessed seismic activity data and corresponding environmental data. Using preprocessed seismic activity data and corresponding environmental data as data samples, a model based on a time-recurrent neural network is trained to obtain a trained seismic activity parameter prediction model.

[0036] In this embodiment, to train the seismic activity parameter prediction model, earthquake-related data and corresponding environmental data of the target area over a certain period of time are first collected. The earthquake-related data are then classified and divided into different datasets according to different seismic activity parameter prediction requirements. For the data in each dataset, the corresponding seismic activity parameters are calculated using time windows set according to research needs; time windows support years, months, days, etc. The calculated seismic activity parameters and corresponding environmental data are preprocessed. Methods such as moving average and wavelet transform are used to remove monitoring noise and calculation errors from the seismic activity parameter time series. Anomalies in the seismic activity parameter time series are detected using the isolated forest algorithm, and interpolation averaging is used to correct these anomalies. Trend features, periodic features, and abrupt change features of the seismic activity parameter time series are extracted. Simultaneously, all parameters are normalized and standardized. The preprocessed dataset is then used as a data sample and input into the model built based on a time-recurrent neural network for training. The model's feature fusion layer employs a weighted fusion mechanism, assigning learnable dynamic weights to different features to integrate multi-source and multi-modal earthquake-related features, uncovering the intrinsic correlations between features, and achieving complementary fusion of multi-source information. Through an improved LSTM layer, optimized input gates, forget gates, candidate memory units, and output gates for seismic activity parameters are introduced, incorporating core seismic activity feature vectors. This effectively captures long-term dependencies in seismic data over time, solving the gradient vanishing or exploding problem, and accurately capturing these dependencies. This aligns with the complex temporal characteristics of seismic activity, such as clustering and stress transfer, making parameter predictions more consistent with the actual patterns of seismic activity. Next, a fully connected layer performs nonlinear transformation and dimensional mapping on the temporal features output from the LSTM layer, employing ReLU. The activation function learns complex nonlinear relationships, and a dynamic dropout rate based on the activation values ​​of seismic activity parameters is introduced. This effectively prevents model overfitting while extracting key features, enhancing the model's generalization ability and stability. Finally, the low-dimensional features output by the fully connected layer are mapped to interpretable seismic activity parameter prediction results through the output layer. The prediction results include predicted seismic activity parameters for a certain period of time in the future, including the predicted maximum magnitude, earthquake frequency, probability of anomalous events, and specific values ​​and trends of seismic activity parameters such as b-value, a-value, λ-value, and P-value. These results can be directly used in applications such as earthquake risk assessment, monitoring and early warning, and seismic analysis of engineering projects.

[0037] The seismic activity parameter prediction model provided in this embodiment, through improvements to the input gate, forget gate, and output gate, not only increases the density of seismic-related feature information, resulting in faster model convergence and more efficient gradient updates, achieving convergence without excessive iterations, but also significantly improves the model's ability to predict seismic activity parameters, increasing the accuracy of predictions by 20% to 50%. By setting a dynamic dropout rate, the model's training convergence can be accelerated, and its stability improved. Under the same hardware environment, the number of iterations required for training is reduced by 25%-30%, and the training time is shortened by 20%-28%.

[0038] In some embodiments, during the model training phase, the predictive performance of the model is calculated in real time by calculating both accuracy and recall. If the preset threshold is not reached, the loss information is automatically transmitted to each layer of the model through the backpropagation algorithm to guide the optimization and adjustment of the weight matrix and bias terms, so that the model can continuously learn and improve its ability to process tasks.

[0039] The accuracy calculation formula is as follows: (5) In the formula, A This indicates the accuracy of the model's prediction results. N c This represents all samples that were predicted correctly. N t This represents all samples.

[0040] The formula for calculating recall rate is: (6) In the formula, R Recall rate represents the model's recall rate. TP This represents the number of samples that are actually positive and are predicted as positive by the model. FN This represents the number of samples that are actually positive but are predicted as negative by the model.

[0041] The seismic activity parameter prediction method provided in this application integrates multi-source data, including earthquake-related data and environmental data. It utilizes a seismic activity parameter prediction model to fuse these multi-source data, fully exploring the intrinsic correlation between environmental data and seismic activity. The parameter analysis more closely aligns with actual geological movement patterns. The model is improved based on the seismic activity parameter prediction task, enhancing training efficiency and increasing the accuracy of prediction results. By configuring time windows, it can meet the research scenarios of seismic activity parameters at different time scales (short, medium, and long term), satisfying the practical application needs of various fields such as earthquake monitoring and early warning, and seismic design of engineering projects.

[0042] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method described.

[0043] It should be noted that the above description describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims may be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0044] like Figure 2 As shown in the illustration, this application also provides a seismic activity parameter prediction device, comprising: The acquisition module is used to acquire seismic activity data and environmental data; The preprocessing module is used to preprocess seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data; The fusion module is used to fuse preprocessed seismic activity data and environmental data to obtain fused features; The prediction module is used to input fused features into a seismic activity parameter prediction model constructed based on a time recurrent neural network, and the seismic activity parameter prediction model outputs the predicted seismic activity parameters.

[0045] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware.

[0046] The apparatus described above is used to implement the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0047] Figure 3 This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0048] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0049] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0050] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0051] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0052] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0053] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0054] The electronic devices described above are used to implement the corresponding methods in the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0055] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0056] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0057] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0058] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0059] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this disclosure.

Claims

1. A method for predicting seismic activity parameters, characterized in that, include: Acquire seismic activity data and environmental data; The seismic activity data and environmental data are preprocessed to obtain preprocessed seismic activity data and environmental data. The preprocessed seismic activity data and environmental data are fused to obtain fused features; The fused features are input into a seismic activity parameter prediction model constructed based on a time-recurrent neural network, and the seismic activity parameter prediction model outputs the predicted seismic activity parameters.

2. The method according to claim 1, characterized in that, The seismic activity parameter prediction model includes an input gate, a forget gate, and an output gate, wherein the input gate is represented as: （1） in, i t This represents the input gate output at the current time step t, where σ represents the activation function. w i This represents the weight matrix of the input gate. b i This represents the bias term of the input gate; h t-1 This indicates the hidden state of the previous time step. P t This represents the seismic activity parameter above a set threshold. ΔT t Anomaly is the time interval between the seismic event detected at the current time step t that is above a set threshold and the last seismic event that was above the set threshold. t This is the seismic anomaly index for the current time step t; The forget gate is represented as follows: （2） in, f t This represents the output of the forget gate at the current time step t. w f The weight matrix representing the forget gate, b f The bias term representing the forget gate; The output gate is represented as: （3） in, o t This indicates the output of the gate at the current time step t. w o This represents the weight matrix of the output gate. b o This represents the bias term of the output gate; The candidate memory units of the model are: （4） in, C t This represents the candidate memory cell at the current time step t. w c The weight matrix represents the candidate memory units. b c The bias term represents the candidate memory unit.

3. The method according to claim 1, characterized in that, The seismic activity parameter prediction model includes a Dropout layer. The Dropout rate of the Dropout layer is determined based on the activation value of the activation function of the previous layer. When the activation value is greater than the mean of the activation values, the Dropout rate is the first probability. When the activation value is less than the mean of the activation values, the Dropout rate is the second probability.

4. The method according to claim 3, characterized in that, The first probability is less than the second probability.

5. The method according to claim 1, characterized in that, Also includes: Collect earthquake-related data and corresponding environmental data for the target area within a certain period of time; Earthquake-related data and corresponding environmental data are divided according to a preset time window to obtain the divided earthquake-related data and corresponding environmental data. Based on the segmented earthquake correlation data, earthquake activity parameters are calculated; The calculated seismic activity parameters and corresponding environmental data are preprocessed to obtain preprocessed seismic activity data and corresponding environmental data. Using the preprocessed seismic activity data and corresponding environmental data as data samples, the model constructed based on a time-recurrent neural network is trained to obtain a trained seismic activity parameter prediction model.

6. A device for predicting seismic activity parameters, characterized in that, include: The acquisition module is used to acquire seismic activity data and environmental data; The preprocessing module is used to preprocess the seismic activity data and environmental data to obtain preprocessed seismic activity data and environmental data. The fusion module is used to fuse the preprocessed seismic activity data and environmental data to obtain fused features; The prediction module is used to input the fused features into a seismic activity parameter prediction model constructed based on a time recurrent neural network, and the seismic activity parameter prediction model outputs the predicted seismic activity parameters.

7. The apparatus according to claim 6, characterized in that, The seismic activity parameter prediction model includes an input gate, a forget gate, and an output gate, wherein the input gate is represented as: （1） in, i t This represents the input gate output at the current time step t, where σ represents the activation function. w i This represents the weight matrix of the input gate. b i This represents the bias term of the input gate; h t-1 This indicates the hidden state of the previous time step. P t This represents the seismic activity parameter above a set threshold. ΔT t Anomaly is the time interval between the seismic event detected at the current time step t that is above a set threshold and the last seismic event that was above the set threshold. t This is the seismic anomaly index for the current time step t; The forget gate is represented as follows: （2） in, f t This represents the output of the forget gate at the current time step t. w f The weight matrix representing the forget gate, b f The bias term representing the forget gate; The output gate is represented as: （3） in, o t This indicates the output of the gate at the current time step t. w o This represents the weight matrix of the output gate. b o This represents the bias term of the output gate; The candidate memory units of the model are: （4） in, C t This represents the candidate memory cell at the current time step t. w c The weight matrix represents the candidate memory units. b c The bias term represents the candidate memory unit.

8. The apparatus according to claim 6, characterized in that, The seismic activity parameter prediction model includes a Dropout layer. The Dropout rate of the Dropout layer is determined based on the activation value of the activation function of the previous layer. When the activation value is greater than the mean of the activation values, the Dropout rate is the first probability. When the activation value is less than the mean of the activation values, the Dropout rate is the second probability.

9. The apparatus according to claim 8, characterized in that, The first probability is less than the second probability.

10. The apparatus according to claim 6, characterized in that, Also includes: The acquisition module is used to collect earthquake-related data and corresponding environmental data of the target area within a certain period of time. The segmentation module is used to segment earthquake-related data and corresponding environmental data according to a preset time window, so as to obtain the segmented earthquake-related data and corresponding environmental data. The calculation module is used to calculate seismic activity parameters based on the segmented seismic correlation data; The preprocessing module is used to preprocess the calculated seismic activity parameters and corresponding environmental data to obtain preprocessed seismic activity data and corresponding environmental data. The training module is used to train the model based on the time recurrent neural network using the preprocessed seismic activity data and corresponding environmental data as data samples, so as to obtain the trained seismic activity parameter prediction model.

Images