Earthquake type identification method and device

Abstract

The invention provides an earthquake type identification method and device, and relates to the technical field of earthquake classification, and the method comprises the steps: firstly extracting the target features of earthquake data after obtaining the to-be-identified earthquake data; wherein the target features are signal features of differences between natural earthquakes and non-natural earthquakes, then processing the target features by using a target Gaussian naive Bayesian network model to obtain an estimated probability that the seismic data belong to the natural earthquakes, and finally calibrating the estimated probability by using a target calibration function to obtain the seismic data of the natural earthquakes. And obtaining a target probability that the seismic data belongs to a natural earthquake, and determining an earthquake type to which the seismic data belongs based on the target probability. In view of the fact that the target calibration function is a function fitted by taking the minimum sum of squares of deviations between the estimated probability mean value of a plurality of target feature sample groups and the real seismic probability as a target, it can be ensured that the seismic type determined by the method has high accuracy and confidence.

Description

Technical Field

[0001] The present invention relates to the technical field of earthquake classification, and in particular to a method and device for identifying earthquake types. Background Art

[0002] With the establishment of dense earthquake monitoring networks around the world and in various countries, various non-natural vibration events will be recorded. These records will be confused with natural earthquake records. If they are not removed in time, the earthquake catalog will be confused and affect the subsequent earthquake hazard analysis. Therefore, solving the earthquake classification problem in regional and local earthquake records is an important task in earthquake monitoring in the future.

[0003] Theoretically, there are great differences in the focal mechanisms of natural earthquakes and artificial explosions. However, due to the large spacing between fixed network stations, it is still difficult to solve the focal parameters of low-magnitude earthquakes (M < 3.0) and the regional high-frequency attenuation model using local and regional seismic records. Therefore, artificial intelligence technologies such as machine learning have been widely used in earthquake and explosion identification. However, in the existing technology in machine learning scenarios, especially in classification tasks, the probability output by the model is the probability that the model predicts that the sample belongs to the positive or negative class based on the training set and model framework, and there may be a large difference between the model prediction result and the actual type attribute of the event. In summary, in the existing technology, the earthquake type identification method has technical problems such as low prediction accuracy and poor credibility. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for identifying earthquake types, so as to alleviate the technical problems of low prediction accuracy and poor credibility of the existing earthquake type identification methods.

[0005] In a first aspect, the present invention provides a method for identifying earthquake types, comprising: acquiring earthquake data to be identified; extracting target features of the earthquake data; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes; processing the target features using a target Gaussian naive Bayes network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake; calibrating the estimated probability using a target calibration function to obtain a target probability that the earthquake data belongs to a natural earthquake, so as to determine the earthquake type to which the earthquake data belongs based on the target probability; wherein the target calibration function is a function fitted with the goal of minimizing the sum of squared deviations between the mean of the estimated probabilities of multiple target feature sample groups and the true earthquake probability.

[0006] Optionally, the target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

[0007] Optionally, before using the target Gaussian naive Bayesian network model to process the target feature, it also includes: obtaining an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a real earthquake type label; if the earthquake data sample belongs to a natural earthquake, the real earthquake type label is 0, otherwise it is 1; extracting the target feature of each earthquake data sample in the earthquake data sample set to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label; using the target feature set to iteratively train the initial Gaussian naive Bayesian network model until a preset end condition is reached to obtain a target Gaussian naive Bayesian network model.

[0008] Optionally, before calibrating the estimated probability using the target calibration function, the method further includes: obtaining a target feature sample set; wherein each target feature sample in the target feature sample set has a real earthquake type label; using a target Gaussian naive Bayesian network model to process each target feature sample in the target feature sample set to determine the estimated probability that each target feature sample belongs to a natural earthquake; dividing the target feature sample set into a specified number of target feature sample groups according to the value of the estimated probability, and calculating the estimated probability mean and the real earthquake probability of each target feature sample group to obtain multiple groups of probability calibration samples; wherein, The estimated probability corresponding to each target feature sample group is in the same value range; the true earthquake probability represents the proportion of the true earthquake type label in the target feature sample group being natural earthquakes; each group of probability calibration samples includes: an estimated probability mean and a true earthquake probability; based on the size of the estimated probability mean, the multiple groups of probability calibration samples are sorted in ascending order to obtain a sorted probability calibration sample group; the sorted probability calibration sample group is fitted using a monotonic regression algorithm to obtain a target calibration function that minimizes the objective function; wherein the objective function is the sum of squared deviations of multiple estimated probability means and corresponding true earthquake probabilities.

[0009] Optionally, the objective function is expressed as: Among them, w i represents the weight, i = 1, 2, ..., n, n represents the number of probability calibration sample groups, y irepresents the true earthquake probability in the i-th group of probability calibration samples, x i represents the estimated probability mean in the i-th group of probability calibration samples.

[0010] In a second aspect, the present invention provides an earthquake type identification device, comprising: an acquisition module for acquiring earthquake data to be identified; an extraction module for extracting target features of the earthquake data; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes; a processing module for processing the target features using a target Gaussian naive Bayes network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake; a calibration module for calibrating the estimated probability using a target calibration function to obtain a target probability that the earthquake data belongs to a natural earthquake, so as to determine the earthquake type to which the earthquake data belongs based on the target probability; wherein the target calibration function is a function fitted with the goal of minimizing the sum of squared deviations between the mean of the estimated probability of multiple target feature sample groups and the true earthquake probability.

[0011] Optionally, the target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

[0012] Optionally, the earthquake type identification device is also used to: obtain an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a real earthquake type label; if the earthquake data sample belongs to a natural earthquake, the real earthquake type label is 0, otherwise it is 1; extract the target features of each earthquake data sample in the earthquake data sample set to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label; use the target feature set to iteratively train the initial Gaussian naive Bayes network model until a preset end condition is reached to obtain a target Gaussian naive Bayes network model.

[0013] In a third aspect, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the processor executes the computer program, the earthquake type identification method of any one of the aforementioned implementation modes is implemented.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the earthquake type identification method of any one of the aforementioned implementation modes is implemented.

[0015] The present invention provides an earthquake type identification method. After obtaining earthquake data to be identified, the target features of the earthquake data are first extracted; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes, and then the target features are processed using a target Gaussian naive Bayesian network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake, and finally the estimated probability is calibrated using a target calibration function to obtain a target probability that the earthquake data belongs to a natural earthquake, and the earthquake type to which the earthquake data belongs is determined based on the target probability. Since the target calibration function is a function fitted with the goal of minimizing the sum of squares of the deviations between the estimated probability mean of multiple target feature sample groups and the true earthquake probability, it can be ensured that the earthquake type determined using the method of the present invention has high accuracy and confidence. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A flowchart of a method for identifying earthquake types provided by an embodiment of the present invention;

[0018] Figure 2 A schematic diagram of 36 characteristic distributions of natural earthquakes and non-natural earthquakes provided in an embodiment of the present invention;

[0019] Figure 3 Schematic diagram of the ROC curve predicted by the target Gaussian Naive Bayesian network model;

[0020] Figure 4 Schematic diagram of the prediction confusion matrix result of the target Gaussian Naive Bayes network model;

[0021] Figure 5 A comparison diagram of earthquake classification effects between the method of the embodiment of the present invention and two comparative methods;

[0022] Figure 6 A functional module diagram of an earthquake type identification device provided by an embodiment of the present invention;

[0023] Figure 7 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0027] Embodiment 1

[0028] Figure 1 A flowchart of a method for identifying earthquake types provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method specifically comprises the following steps:

[0029] Step S102, obtaining earthquake data to be identified.

[0030] Step S104, extracting target features of seismic data.

[0031] Among them, the target feature is the signal feature that is different between natural earthquakes and non-natural earthquakes.

[0032] Specifically, the seismic data to be identified is the seismic record data recorded by the seismic monitoring station. After acquiring the seismic data, the seismic data is subjected to instrument response removal, trend removal, filtering, removal of abnormal records and waveform data with low signal-to-noise ratio in combination with digital signal processing methods and seismological theory, and then the target features are extracted therefrom. In the embodiment of the present invention, in order to assist in the identification of natural earthquake events and non-natural earthquake events, the target features select signal features that differ between natural earthquakes and non-natural earthquakes to highlight the real earthquake type. The target features include: time domain features and frequency domain features. The embodiment of the present invention does not specifically limit the type and quantity of target features, and users can set them according to actual needs.

[0033] Step S106, using the target Gaussian Naive Bayesian network model to process the target features, to obtain the estimated probability that the earthquake data belongs to a natural earthquake.

[0034] Specifically, the sklearn library can be used to construct an initial Gaussian naive Bayesian network model, train the initial Gaussian naive Bayesian network model, and continuously optimize its model parameters until the preset performance index requirements are met, and the target Gaussian naive Bayesian network model can be obtained. The input data of the target Gaussian naive Bayesian network model is the target feature, and the output data is the estimated probability that the earthquake data corresponding to the target feature belongs to a natural earthquake. The sum of the estimated probability of a natural earthquake and the estimated probability of a non-natural earthquake is 1. If the estimated probability of a natural earthquake is greater than the estimated probability of a non-natural earthquake, it means that the target Gaussian naive Bayesian network model preliminarily determines that the earthquake type to which the earthquake data belongs is a natural earthquake. However, in view of the low accuracy and credibility of the target Gaussian naive Bayesian network model, the embodiment of the present invention further needs to calibrate the estimated probability output by the target Gaussian naive Bayesian network model.

[0035] Step S108, calibrating the estimated probability using the target calibration function to obtain the target probability that the seismic data belongs to a natural earthquake, so as to determine the earthquake type to which the seismic data belongs based on the target probability.

[0036] Among them, the target calibration function is a function fitted with the goal of minimizing the sum of squares of the deviations between the estimated probability mean of multiple target feature sample groups and the true earthquake probability.

[0037] That is to say, the input data of the target calibration function is the output data of the target Gaussian naive Bayesian network model. The estimated probability is processed by the target calibration function, and the output probability is the target probability that the earthquake data belongs to a natural earthquake. In view of the fact that the target calibration function is a function fitted with the minimum sum of squares of the deviations between the mean of a large number of estimated probabilities and the true earthquake probability as the goal, the probability calibrated by the target calibration function has a higher confidence and accuracy. After the embodiment of the present invention processes the earthquake data to be identified, the earthquake type identification result finally obtained includes both the earthquake type to which the earthquake data belongs and the uncertainty (that is, probability) of the earthquake type. For example, if the target probability that the earthquake data belongs to a natural earthquake is 80%, the earthquake type identification result is: natural earthquake (80%); if the target probability that the earthquake data belongs to a natural earthquake is 30%, the earthquake type identification result is: non-natural earthquake (70%).

[0038] The embodiment of the present invention provides a method for identifying earthquake types. After obtaining earthquake data to be identified, the target features of the earthquake data are first extracted; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes, and then the target features are processed using a target Gaussian naive Bayesian network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake, and finally the estimated probability is calibrated using a target calibration function to obtain a target probability that the earthquake data belongs to a natural earthquake, and the earthquake type to which the earthquake data belongs is determined based on the target probability. Given that the target calibration function is a function fitted with the goal of minimizing the sum of squared deviations between the estimated probability mean of multiple target feature sample groups and the true earthquake probability, it is possible to ensure that the earthquake type determined using the method of the present invention has a high degree of confidence and accuracy.

[0039] In an optional embodiment, the target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

[0040] The more target features there are, the more stable the earthquake type identification result is. In order to highlight the differences between different types of earthquake events as much as possible, the embodiment of the present invention proposes to quantify the above 8 types of earthquake features, where the user can obtain the P / S amplitude ratio of multiple frequency bands according to actual needs. If the number of frequency bands is 21, correspondingly, Figure 2 A schematic diagram of 36 characteristic distributions of natural earthquakes and non-natural earthquakes provided in an embodiment of the present invention, Figure 2 The artificial explosion in the above data represents a non-natural earthquake. The following is an introduction to the extraction method of the corner frequency and waveform complexity.

[0041] It is known that the corner frequency is the intersection of the high-frequency asymptotic trend and the low-frequency asymptotic trend of the amplitude spectrum, and it is also a physical quantity that reflects the size of the earthquake source. The seismic wave amplitude spectrum is a description of the seismic waves radiated by the earthquake source in the frequency domain. It is closely related to the mechanical parameters of the earthquake source. The theoretical seismic wave amplitude spectra radiated by different earthquake source modes are different. For small and medium earthquakes, the theoretical earthquake source spectrum conforms to the Brune model. When the inelastic attenuation is ignored, the Brune model earthquake source spectrum Ω(f) is expressed as: Ω(f) = Ω0 / (1+(f / f c ) 2 ); the Jimenez model source spectrum Ω(f) is expressed as:

[0042] In the above two expressions, Ω0 represents the zero-frequency limit of the displacement spectrum, f represents the frequency, and f c The embodiment of the present invention uses the difference between the observed amplitude spectrum and the theoretical spectrum as the objective function, and calculates the corner frequency f at the point where the deviation between the theoretical value and the actual value is the smallest. c Referring to the above methods, the P-wave and S-wave corner frequencies can be calculated according to the Brune model and the Jimenez model respectively.

[0043] It is generally believed that the P wave radiated by the explosion source is simpler than the P wave radiated by the natural earthquake source, that is, the energy of the explosion is mainly concentrated in the front part of the time period, and the tail part is suppressed, making the entire waveform relatively simpler than that of a natural earthquake; while a natural earthquake has various seismic phases, and there is energy distribution at the front and back of the time period, and the overall waveform is more complex than the explosion waveform. The formula for defining the waveform complexity in the embodiment of the present invention is: Where C represents the waveform complexity. If the P wave waveform complexity is calculated, t1 and t2 represent the start time and end time of the P wave respectively. If the S wave waveform complexity is calculated, t1 and t2 represent the start time and end time of the S wave respectively. p represents the arrival time of the P wave, t coda It represents the time when the waveform decays to the noise level, and y(t) represents the amplitude of the waveform at time t.

[0044] In an optional implementation manner, before processing the target features using the target Gaussian Naive Bayesian network model, the method of the present invention further includes the following steps:

[0045] Step S201, obtaining an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a real earthquake type label; if the earthquake data sample belongs to a natural earthquake, the real earthquake type label is 0, otherwise it is 1.

[0046] Step S202, extracting the target feature of each earthquake data sample in the earthquake data sample set to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label.

[0047] Step S203, iteratively training the initial Gaussian Naive Bayesian network model using the target feature set until a preset end condition is reached to obtain a target Gaussian Naive Bayesian network model.

[0048] Specifically, before model training, the embodiment of the present invention sets the earthquake data samples of natural earthquakes, whose real earthquake type labels are 0; the earthquake data samples of non-natural earthquakes, whose real earthquake type labels are 1. After obtaining the earthquake data sample set, feature extraction is performed on each sample therein, and then the target features of each sample are processed using the initial Gaussian naive Bayesian network model to predict the probability of it belonging to a natural earthquake, and then the loss function value is calculated in combination with the real earthquake type label of the sample, and then the model is trained using the loss function value. The model training can also be carried out in a batch processing manner. When the model training reaches a preset round, or the accuracy of the model prediction result reaches a specified threshold, the training can be stopped to obtain the target Gaussian naive Bayesian network model.

[0049] Figure 3 The ROC curve diagram for the target Gaussian Naive Bayesian network model prediction is shown in Figure 1. The AUC is 0.9763. Figure 4 This is a schematic diagram of the prediction confusion matrix results of the target Gaussian Naive Bayes network model, in which Class 0 (natural earthquakes) was accurately identified 574 times and Class 1 (non-natural earthquakes) was accurately identified 425 times.

[0050] In an optional implementation manner, before calibrating the estimated probability using the target calibration function, the method of the present invention further includes the following steps:

[0051] Step S301, obtaining a target feature sample set; wherein each target feature sample in the target feature sample set has a real earthquake type label.

[0052] Step S302, using the target Gaussian Naive Bayesian network model to process each target feature sample in the target feature sample set to determine the estimated probability that each target feature sample belongs to a natural earthquake.

[0053] Step S303, divide the target feature sample set into a specified number of target feature sample groups according to the value of the estimated probability, and calculate the estimated probability mean and the true earthquake probability of each target feature sample group to obtain multiple groups of probability calibration samples; wherein the estimated probability corresponding to each target feature sample group is in the same value range; the true earthquake probability represents the proportion of the true earthquake type label in the target feature sample group being natural earthquakes; each group of probability calibration samples includes: the estimated probability mean and the true earthquake probability.

[0054] Step S304 , sorting the multiple groups of probability calibration samples in ascending order based on the size of the estimated probability mean value to obtain a sorted probability calibration sample group.

[0055] Step S305, using a monotonic regression algorithm to fit the sorted probability calibration sample group to obtain a target calibration function that minimizes the target function; wherein the target function is the sum of squares of deviations between multiple estimated probability means and corresponding true earthquake probabilities.

[0056] The above steps S301 to S305 are the method flow for determining the target calibration function. Specifically, after obtaining the target Gaussian naive Bayesian network model through training, the model is used to predict the earthquake type for each sample feature in the target feature sample set (that is, the target feature sample) to obtain the estimated probability that the sample feature belongs to a natural earthquake.

[0057] Next, on the condition that the estimated probability corresponding to each target feature sample group is in the same value range, the target feature sample set is divided into a specified number of target feature sample groups according to the value of the estimated probability. For example, the estimated probability is equally spaced. For example, the target feature sample set is divided into 10 target feature sample groups at an interval of 0.1, that is, the estimated probability (0, 0.1) is a group, (0.1, 0.2) is a group, and so on. According to the value of the estimated probability, a total of 10 target feature sample groups are divided.

[0058] Next, the estimated probability mean and true earthquake probability of each target feature sample group are calculated to obtain multiple groups of probability calibration samples. If there are 100 target feature samples in a target feature sample group, and the true earthquake type label of 15 of the target feature samples is natural earthquake, then the true earthquake probability of the target feature sample group is 0.15. The estimated probability mean is the average value of the estimated probabilities of the above 100 target feature samples.

[0059] Then, multiple groups of probability calibration samples are sorted in order from small to large according to the estimated probability mean value. After sorting, the monotonic regression algorithm is used to fit them. When the objective function is minimized, the target calibration function is obtained, which is also the correction curve expression of the estimated probability.

[0060] In an optional implementation, the objective function is expressed as: Among them, w i represents the weight, i = 1, 2, ..., n, n represents the number of probability calibration sample groups, y i represents the true earthquake probability in the i-th group of probability calibration samples, x i represents the estimated probability mean in the i-th group of probability calibration samples.

[0061] In order to verify the performance of the method provided in the embodiment of the present invention, the method provided in the embodiment of the present invention is compared with the following two earthquake type identification methods: a Gaussian naive Bayesian network model, and a method of calibrating the output probability of the Gaussian naive Bayesian network model using the Pratt scaling method (hereinafter referred to as Pratt scaling calibration). Each method draws an earthquake classification effect curve with reference to the following process.

[0062] ① Use earthquake type identification method to calculate the predicted probability value of all target characteristic samples;

[0063] ② Divide the predicted probability values ​​into several groups;

[0064] ③ Calculate the mean of each group of predicted probability values, that is, an estimated probability mean;

[0065] ④ Calculate the proportion of each group of positive classes to obtain the true earthquake probability;

[0066] ⑤ Use the mean of each group of estimated probability as the horizontal axis and the true earthquake probability of each group as the vertical axis to draw a scatter plot and connect the points.

[0067] Figure 5 The following is a comparison chart of earthquake classification effects between the method of the embodiment of the present invention and two comparative methods. Figure 5 The dotted line in the earthquake classification effect curve is a diagonal line. The predicted value of each point on the diagonal line is equal to the true value, that is, the mean predicted probability of each group is equal to the true earthquake probability of the group. In other words, the diagonal line is a curve with completely correct classification. The closer the classification effect curve of the earthquake type recognition method is to the diagonal line, the better the calibration effect is.

[0068] pass Figure 5 It can be seen that the uncalibrated Gaussian Naive Bayes model has a large oscillation curve and a poor prediction probability effect. The reason is that the extracted features violate the assumption of Bayesian independence distribution, resulting in the classifier's confidence being too confident. The Pratt scaling calibrated curve also oscillates greatly, and the improvement is small compared to the uncalibrated Gaussian Naive Bayes model, and through Figure 5 As can be seen from the histogram in , the extracted features violate the assumption of Bayesian independence distribution, and the classification effect is poor. Compared with the above two methods, the method provided by the embodiment of the present invention has a smaller curve oscillation and is close to the diagonal line. Obviously, the method of the present invention improves the calibration result well.

[0069] Furthermore, multiple comprehensive quantitative indicators are used to analyze the advantages and disadvantages of the earthquake classification effects of the above three methods, including Brier score, log loss, precision, recall, F1score, ROC and AUC. The following Table 1 shows the comparison results of the comprehensive evaluation indicators of the three methods.

[0070] Table 1

[0071]

[0072] It can be seen from Table 1 that the three earthquake type identification methods are close in terms of precision, recall rate and F1 score. In terms of logarithmic loss and Brier score indicators, the uncalibrated Gaussian naive Bayes model has the highest score, indicating that the predicted probability of the model is the most different from the actual result. The predicted probability of the Pratt scaling model is improved after calibration, while the method provided by the embodiment of the present invention has the lowest score, indicating that this method is the best, the predicted probability is most consistent with the actual result, and the calibration effect has reached the expected result.

[0073] Embodiment 2

[0074] The embodiment of the present invention further provides an earthquake type identification device, which is mainly used to execute the earthquake type identification method provided in the above-mentioned embodiment 1. The earthquake type identification device provided in the embodiment of the present invention is specifically introduced below.

[0075] Figure 6 A functional module diagram of an earthquake type identification device provided by an embodiment of the present invention, such as Figure 6 As shown, the device mainly includes: an acquisition module 10, an extraction module 20, a processing module 30, and a calibration module 40, wherein:

[0076] The acquisition module 10 is used to acquire the seismic data to be identified.

[0077] The extraction module 20 is used to extract target features of seismic data; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes.

[0078] The processing module 30 is used to process the target features using the target Gaussian Naive Bayesian network model to obtain the estimated probability that the earthquake data belongs to a natural earthquake.

[0079] The calibration module 40 is used to calibrate the estimated probability using the target calibration function to obtain the target probability that the seismic data belongs to a natural earthquake, so as to determine the earthquake type to which the seismic data belongs based on the target probability; wherein the target calibration function is a function fitted with the goal of minimizing the sum of squares of the deviations between the mean of the estimated probability of multiple target feature sample groups and the true earthquake probability.

[0080] The embodiment of the present invention provides an earthquake type identification device. After obtaining the earthquake data to be identified, the target feature of the earthquake data is first extracted; wherein the target feature is a signal feature that is different between natural earthquakes and non-natural earthquakes, and then the target feature is processed using a target Gaussian naive Bayesian network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake, and finally the estimated probability is calibrated using a target calibration function to obtain a target probability that the earthquake data belongs to a natural earthquake, and the earthquake type to which the earthquake data belongs is determined based on the target probability. Since the target calibration function is a function fitted with the goal of minimizing the sum of squares of the deviations between the estimated probability mean of multiple target feature sample groups and the true earthquake probability, it can ensure that the earthquake type determined using the device of the present invention has high accuracy and confidence.

[0081] Optionally, the target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

[0082] Optionally, the earthquake type identification device is further used for:

[0083] Obtain an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a true earthquake type label; if the earthquake data sample belongs to a natural earthquake, the true earthquake type label is 0, otherwise it is 1.

[0084] The target feature of each earthquake data sample in the earthquake data sample set is extracted to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label.

[0085] The initial Gaussian Naive Bayesian network model is iteratively trained using the target feature set until the preset end condition is reached to obtain the target Gaussian Naive Bayesian network model.

[0086] Optionally, the earthquake type identification device is further used for:

[0087] A target feature sample set is obtained; wherein each target feature sample in the target feature sample set has a real earthquake type label.

[0088] The target Gaussian Naive Bayesian network model is used to process each target feature sample in the target feature sample set to determine the estimated probability that each target feature sample belongs to a natural earthquake.

[0089] The target feature sample set is divided into a specified number of target feature sample groups according to the value of the estimated probability, and the estimated probability mean and the true earthquake probability of each target feature sample group are calculated to obtain multiple groups of probability calibration samples; wherein the estimated probability corresponding to each target feature sample group is in the same value range; the true earthquake probability represents the proportion of the true earthquake type label in the target feature sample group being natural earthquakes; each group of probability calibration samples includes: the estimated probability mean and the true earthquake probability.

[0090] The multiple groups of probability calibration samples are sorted in ascending order based on the size of the estimated probability mean value to obtain a sorted probability calibration sample group.

[0091] The sorted probability calibration sample group is fitted using a monotonic regression algorithm to obtain a target calibration function that minimizes the target function; wherein the target function is the sum of squares of deviations between multiple estimated probability means and corresponding true earthquake probabilities.

[0092] Optionally, the objective function is expressed as: Among them, w i represents the weight, i = 1, 2, ..., n, n represents the number of probability calibration sample groups, y i represents the true earthquake probability in the i-th group of probability calibration samples, x i represents the estimated probability mean in the i-th group of probability calibration samples.

[0093] Embodiment 3

[0094] See also Figure 7 An embodiment of the present invention provides an electronic device, which includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, the processor 60, the communication interface 63 and the memory 61 are connected through the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0095] The memory 61 may include a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0096] The bus 62 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, Figure 7Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0097] Among them, the memory 61 is used to store programs, and the processor 60 executes the program after receiving the execution instruction. The method executed by the device defined by the process disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0098] The processor 60 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 60. The above processor 60 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0099] A computer program product of an earthquake type identification method and device provided in an embodiment of the present invention includes a computer-readable storage medium storing a non-volatile program code executable by a processor. The instructions included in the program code can be used to execute the method in the previous method embodiment. The specific implementation can be found in the method embodiment, which will not be repeated here.

[0100] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0101] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0102] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0103] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0104] In addition, the terms "horizontal", "vertical", "overhanging" and the like do not mean that the components are required to be absolutely horizontal or overhanging, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0105] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying earthquake types, characterized in that: include: Acquiring earthquake data to be identified; Extracting target features of the seismic data; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes; Processing the target features using a target Gaussian Naive Bayesian network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake; The estimated probability is calibrated using a target calibration function to obtain a target probability that the seismic data belongs to a natural earthquake, so as to determine the earthquake type to which the seismic data belongs based on the target probability; wherein the target calibration function is a function fitted with the goal of minimizing the sum of squares of deviations between the mean of the estimated probability of multiple target feature sample groups and the true earthquake probability.

2. The earthquake type identification method according to claim 1, characterized in that: The target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

3. The earthquake type identification method according to claim 1, characterized in that: Before processing the target feature using the target Gaussian Naive Bayesian network model, the method further includes: Acquire an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a real earthquake type label; if the earthquake data sample belongs to a natural earthquake, the real earthquake type label is 0, otherwise it is 1; Extracting the target feature of each seismic data sample in the seismic data sample set to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label; The initial Gaussian Naive Bayesian network model is iteratively trained using the target feature set until a preset end condition is reached, thereby obtaining the target Gaussian Naive Bayesian network model.

4. The earthquake type identification method according to claim 1, characterized in that: Before calibrating the estimated probability using the target calibration function, the method further includes: Acquire a target feature sample set; wherein each target feature sample in the target feature sample set has a real earthquake type label; Using the target Gaussian Naive Bayesian network model to process each target feature sample in the target feature sample set to determine the estimated probability that each target feature sample belongs to a natural earthquake; The target feature sample set is divided into a specified number of target feature sample groups according to the value of the estimated probability, and the estimated probability mean and the true earthquake probability of each target feature sample group are calculated to obtain multiple groups of probability calibration samples; wherein the estimated probability corresponding to each target feature sample group is in the same value range; the true earthquake probability represents the proportion of the true earthquake type label in the target feature sample group being a natural earthquake; each group of probability calibration samples includes: the estimated probability mean and the true earthquake probability; Sorting the multiple groups of probability calibration samples in ascending order based on the size of the estimated probability means to obtain a sorted probability calibration sample group; The sorted probability calibration sample group is fitted using a monotonic regression algorithm to obtain a target calibration function that minimizes the target function; wherein the target function is the sum of squares of deviations between a plurality of estimated probability means and corresponding true earthquake probabilities.

5. The earthquake type identification method according to claim 4, characterized in that: The objective function is expressed as: Among them, w i represents the weight, i = 1, 2, ..., n, n represents the number of probability calibration sample groups, y i represents the true earthquake probability in the i-th group of probability calibration samples, x i Represents the estimated probability mean in the i-th group of probability calibration samples.

6. An earthquake type identification device, characterized in that: include: An acquisition module, used for acquiring earthquake data to be identified; An extraction module, used to extract target features of the seismic data; wherein the target features are signal features that differ between natural earthquakes and non-natural earthquakes; A processing module, used to process the target feature using a target Gaussian naive Bayesian network model to obtain an estimated probability that the earthquake data belongs to a natural earthquake; A calibration module is used to calibrate the estimated probability using a target calibration function to obtain a target probability that the seismic data belongs to a natural earthquake, so as to determine the earthquake type to which the seismic data belongs based on the target probability; wherein the target calibration function is a function fitted with the goal of minimizing the sum of squares of deviations between the mean of the estimated probability of multiple target feature sample groups and the true earthquake probability.

7. The earthquake type identification device according to claim 6, characterized in that: The target features include at least one of the following: P / S amplitude ratio of a specified frequency band, P-wave high-to-low frequency energy ratio, S-wave high-to-low frequency energy ratio, P-wave corner frequency based on Brune source spectrum model, S-wave corner frequency based on Brune source spectrum model, P-wave corner frequency based on Jimenez source spectrum model, S-wave corner frequency based on Jimenez source spectrum model, waveform duration, P-wave waveform complexity, S-wave waveform complexity, P-wave zero crossing rate, S-wave zero crossing rate, P-wave cepstrum complexity, S-wave cepstrum complexity, P-wave instantaneous frequency complexity, and S-wave instantaneous frequency complexity.

8. The earthquake type identification device according to claim 6, characterized in that: The earthquake type identification device is also used for: Acquire an earthquake data sample set; wherein each earthquake data sample in the earthquake data sample set has a real earthquake type label; if the earthquake data sample belongs to a natural earthquake, the real earthquake type label is 0, otherwise it is 1; Extracting the target feature of each seismic data sample in the seismic data sample set to obtain a target feature set; wherein each target feature in the target feature set corresponds to a real earthquake type label; The initial Gaussian Naive Bayesian network model is iteratively trained using the target feature set until a preset end condition is reached, thereby obtaining the target Gaussian Naive Bayesian network model.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the computer program, the earthquake type identification method described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the earthquake type identification method according to any one of claims 1 to 5 is implemented.

Images