Event monitoring method and device

Through the Gaussian process regression model and Poisson distribution, a time-varying frequency curve is constructed, which solves the problem of low event monitoring efficiency in the existing technology, realizes rapid identification of high-frequency incident intervals, and improves monitoring efficiency.

CN112732519BActive Publication Date: 2024-12-03BIGO TECH PTE LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011611537.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-30
Publication Date
2024-12-03
Estimated Expiration
2040-12-30

AI Technical Summary

Technical Problem

The existing event monitoring system is difficult to quickly identify the high-frequency range of events when processing large amounts of data, resulting in inefficient monitoring.

Method used

The Gaussian process regression model is used to combine Poisson distribution to construct a time-varying frequency curve, determine the frequency of monitoring events through historical occurrence time information, and generate a frequency curve reflecting the time changes to identify high-frequency generation intervals.

Benefits of technology

It improves the recognition efficiency of monitoring events, especially suitable for scenarios with large data volumes, can quickly identify the high-frequency range of the current monitoring events, and improves the response capability of the monitoring system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112732519B_ABST
    Figure CN112732519B_ABST
Patent Text Reader

Abstract

The present application discloses an event monitoring method and device, the method comprising: obtaining historical occurrence data of a current monitoring event within a specified historical time interval, the historical occurrence data comprising historical occurrence time information of the monitoring event; constructing a Gaussian process regression model according to the historical occurrence time information; determining the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model, wherein the occurrence frequency obeys a Poisson distribution; generating a time-varying frequency curve for the monitoring event according to the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval, the time-varying frequency curve being used to reflect the change in the occurrence frequency of the monitoring event over time, thereby being able to quickly identify the high-frequency occurrence interval of the current monitoring event, being particularly suitable for scenarios with relatively large amounts of data, and improving the recognition efficiency of monitoring events.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to data processing technology, and more particularly to an event monitoring method and device. Background Art

[0002] The monitoring system is of great significance to the operation and maintenance management and risk control of the production system. Among them, event monitoring is a relatively basic and important content. By monitoring special events or important events, the frequency of occurrence of these events can be obtained, so that the corresponding information can be mined or timely responses can be made when events occur.

[0003] For example, in live broadcast scenarios, various events often occur, such as: someone gets a special reward with a low probability in a small game on the live broadcast platform, the number of people in the live broadcast room increases or decreases sharply, etc. It is meaningful for the live broadcast platform to monitor the changes in the frequency of these events. For example, for the event that someone gets a special reward with a low probability in a small game on the live broadcast platform, if the event occurs too frequently, timely response can quickly detect cheating players. Summary of the invention

[0004] The present application provides an event monitoring method and device to determine an interval in which a monitoring event occurs more frequently and to improve the recognition efficiency of the monitoring event.

[0005] In a first aspect, an embodiment of the present application provides an event monitoring method, the method comprising:

[0006] Obtaining historical occurrence data of a current monitoring event within a specified historical time interval, wherein the historical occurrence data includes historical occurrence time information of the monitoring event;

[0007] Constructing a Gaussian process regression model based on the historical occurrence time information;

[0008] Determine, according to the Gaussian process regression model, the frequency of occurrence of the monitoring event at each unit time point within the specified historical time interval, wherein the frequency of occurrence obeys a Poisson distribution;

[0009] According to the occurrence frequency of the monitoring event at each unit time point in the specified historical time interval, a time-varying frequency curve for the monitoring event is generated, and the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time.

[0010] In a second aspect, an embodiment of the present application further provides an event monitoring device, the device comprising:

[0011] A historical data acquisition module is used to acquire historical occurrence data of a current monitoring event within a specified historical time interval, wherein the historical occurrence data includes historical occurrence time information of the monitoring event;

[0012] A model building module, used to build a Gaussian process regression model according to the historical occurrence time information;

[0013] An occurrence frequency determination module, used to determine the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model, wherein the occurrence frequency obeys a Poisson distribution;

[0014] The frequency curve generation module is used to generate a time-varying frequency curve for the monitoring event according to the occurrence frequency of the monitoring event at each unit time point in the specified historical time interval, and the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time.

[0015] In a third aspect, an embodiment of the present application further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.

[0016] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned method when executed by a processor.

[0017] This application has the following beneficial effects:

[0018] In this embodiment, in the scenario of event monitoring, a Gaussian process regression model can be constructed based on the collected historical occurrence time information of the current monitoring event within the specified historical time interval, and then the occurrence frequency of the current monitoring event at each unit time point within the specified historical time interval can be determined based on the Gaussian process regression model, and a time-varying frequency curve for the monitoring event can be drawn based on the occurrence frequency. Based on the time-varying frequency curve, the change of the occurrence frequency of the current monitoring event over time can be determined, and then the high-frequency occurrence interval of the current monitoring event can be quickly identified, which is particularly suitable for scenarios with relatively large data volumes, and improves the recognition efficiency of monitoring events. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of an event monitoring method embodiment provided in Embodiment 1 of the present application;

[0020] Figure 2 is a flow chart for determining the occurrence frequency of a monitoring event provided in Embodiment 1 of the present application;

[0021] Figure 3 It is a schematic diagram of a time-varying frequency curve provided in Example 1 of the present application;

[0022] Figure 4 This is a schematic diagram of determining a high-frequency transmission interval according to a time-varying frequency curve provided in Embodiment 1 of the present application;

[0023] Figure 5 It is a structural block diagram of an event monitoring device embodiment provided in Embodiment 2 of the present application;

[0024] Figure 6 It is a structural schematic diagram of an electronic device provided in Example 3 of the present application. DETAILED DESCRIPTION

[0025] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only the parts related to the present application, rather than all structures, are shown in the accompanying drawings.

[0026] Embodiment 1

[0027] Figure 1 A flowchart of an event monitoring method embodiment provided in Embodiment 1 of the present application. This embodiment can be executed by a server. For example, the server may include a server of a platform (such as a live broadcast platform, a short video platform, etc.).

[0028] This embodiment may include the following steps:

[0029] Step 110, obtaining historical occurrence data of the current monitoring event within a specified historical time interval, wherein the historical occurrence data includes historical occurrence time information of the monitoring event.

[0030] In one example, monitoring events may include events specified by the platform that need to be monitored or observed. For example, monitoring events may include platform network connection fluctuation events, network failure events, events such as a sudden increase or decrease in the number of people in the live broadcast room, etc. This embodiment does not limit specific monitoring events and can be set according to actual business needs.

[0031] After determining the monitoring event, this embodiment can collect the historical occurrence data of the monitoring event within the specified historical time interval, that is, the data generated each time the monitoring event occurs in the past time period. Exemplarily, the historical occurrence data can at least include the historical occurrence time information of the current monitoring event, for example, the historical occurrence time information can be the timestamp information of each occurrence of the current monitoring event in the past month.

[0032] In one implementation, historical occurrence data of the current monitoring event within a specified historical time interval can be collected from log data. Of course, this embodiment is not limited to the method of collecting historical occurrence data from log data. Those skilled in the art can use other methods to collect historical occurrence data of the current monitoring event within a specified historical time interval. At the same time, this embodiment does not limit the specific specified historical time interval, which can be determined according to actual business needs.

[0033] In one embodiment, the above-mentioned historical occurrence time information can be time information after normalization. Specifically, in order to facilitate the subsequent fitting of the model, after the historical occurrence data of the current monitoring event within the specified historical time interval is collected, the historical occurrence time in the historical occurrence data can be standardized to normalize the time information to a small range. For example, the time information of the past month is normalized to between [0, 2.0]. Then, the subsequent historical occurrence time information or the unit time point within the specified historical time interval are all normalized time information. For example, a form similar to t1=0, t2=0.01, t3=0.02, t4=0.23 is used to record the time point of each occurrence of the monitoring event.

[0034] Step 120: construct a Gaussian process regression model based on the historical occurrence time information.

[0035] In this embodiment, it is assumed that the time interval of the current monitoring event is generated by a Poisson distribution sampling. Specifically, the Poisson process can be used to model the occurrence frequency of the monitoring event. The Poisson process assumes that the number of events occurring in two mutually exclusive (non-overlapping) intervals are independent random variables, and the probability distribution of the number of events occurring in the interval [t, t+τ] is:

[0036]

[0037] Where λ is a positive number, usually called arrival rate or intensity. Then, if a time interval [t, t+τ] is given, the number of events occurring in the time interval is a Poisson distribution with parameter λτ. In other words, in the time interval [t, t+τ], the expectation of the occurrence of the current monitoring event is λτ.

[0038] In the existing Poisson process, λτ is a fixed value, that is, it is pre-assumed that the expected value of an event occurring within a fixed time interval is constant. In this embodiment, it is assumed that λτ is a time-varying parameter, and λτ will change with the change of time, rather than a fixed value. Based on this, this embodiment can construct a Gaussian process regression model based on the historical occurrence time information in the historical data to characterize the change process of λτ.

[0039] Among them, Gaussian Process Regression (GPR) is a non-parameteric model that uses Gaussian Process (GP) prior to perform regression analysis on data. The GPR model assumption includes two parts: noise (regression residual) and Gaussian process prior, and its solution is performed according to Bayesian inference. If the form of the kernel function is not restricted, GPR is theoretically a universal approximator for any continuous function in a compact space. In addition, GPR can provide a posterior for the prediction results, and when the likelihood is a normal distribution, the posterior has an analytical form.

[0040] In one embodiment, step 120 may further include the following steps:

[0041] Step 120 - 1 , taking the historical occurrence time information as a sample set, using the conjugate gradient method to minimize the log likelihood of the sample set, obtaining optimal hyperparameters, and determining a Gaussian process regression model based on the optimal hyperparameters.

[0042] In implementation, according to Bayes' theorem, the following formula is obtained:

[0043]

[0044] Among them, y = (y1, y2, ...) is the historical occurrence time information, y i represents the time when the i-th monitoring event occurs. In this embodiment, That is, the feature space of this embodiment is a positive real number.

[0045] s = (f(y1), f(y2), ...) is the sampling of the Gaussian process regression model f on y;

[0046] p(f)=GP(m,k), where m and k are the mathematical expectation function and covariance function respectively.

[0047] is the likelihood function.

[0048] p(s)=∫p(s|f,y)p(f|s)df is the evidence function.

[0049] In an embodiment, in a given In the case of , the likelihood function of an exemplary Gaussian process regression model can be expressed by the following formula using the conjugate gradient method:

[0050]

[0051] Among them, y1,…,y n is the time variable of the observed n monitoring events, and T is the maximum value of the time variable in the observed data (i.e., historical occurrence time information). is the time at which the monitored event occurs. t The logarithm and is the expected number of monitoring events that occur during the observation data period. The likelihood function can be understood as the expectation that when a monitoring event occurs, t The value of is high. But at the same time, we need to control the average λ t The level of λ makes the monitoring event occur frequently in the interval λ t High, the interval λ where events occur sparsely t The level is low.

[0052] In this embodiment, the optimization method is used to minimize logp(y1, ..., y n |λ), we can get λ t The estimated value is the optimal hyperparameter. After obtaining the optimal hyperparameter, the Gaussian process regression model can be determined. Among them, this embodiment minimizes logp(y1, ..., y n |λ) is not limited, for example, Bayesian variational inference can be used for optimization.

[0053] Step 120 - 2 , according to the Gaussian process regression model, determine the corresponding posterior mathematical expectation function and posterior covariance function to determine the posterior distribution probability.

[0054] Specifically, the goal of Gaussian process regression modeling is to find the posterior distribution probability of f given y and s. That is, the posterior distribution probability is the goal to be found in the task of Gaussian process regression. In one embodiment, the posterior function distribution probability can be set as p(f|s, y) = GP(m post ,k post ), to determine the probability of the posterior function distribution, we need to first find the posterior mathematical expectation function m post and the posterior covariance function k post .

[0055] In one embodiment, the solution of Gaussian process regression has the following analytical form:

[0056] m post (y ★ )=m(y ★ )+k(y ★ ,s)K -1 (sm(y))

[0057] k post (y ★ ,y ★ )=k(y ★ ,y ★ )-k(y ★ ,y)K -1 k(y,y ★ )

[0058] Among them, y ★ is any possible value in the feature space. In this embodiment, y ★ YesR + Any possible number on represents the value of any possible time at which the monitoring event may occur.

[0059] m(y) is the prior distribution of the mathematical expectation function of the Gaussian process on the input sample. In this embodiment, it is the event frequency function in R + The prior distribution of is usually 0.

[0060] K = k(y, y) is the covariance matrix of the input variables in the sample set. In this embodiment, the input variables are historical occurrence time information. -1 is the transpose of K.

[0061] k(y ★ , y) is the covariance vector of the input variables of the data point to be predicted, in this embodiment, it is R + The covariance function value of any possible value of and the historical occurrence time information in history.

[0062] k(y,y ★ ) is k(y ★ , the transpose of y).

[0063] It should be noted that in this embodiment, no specific covariance function is specified. Generally speaking, when the covariance matrix is ​​a semi-positive definite matrix, the kernel function is considered to be valid. When the covariance matrix is ​​a semi-positive definite matrix, the covariance function is considered to be valid. For example, the covariance function can be expressed by the following formula:

[0064] k(x1,x2)=exp(-|d(x1,x2)|2 )

[0065] Where d(x1, x2) is the Euclidean distance between any two sample points x1 and x2.

[0066] Step 130 , determining the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model, wherein the occurrence frequency obeys a Poisson distribution.

[0067] After constructing the Gaussian process regression model, the frequency of monitoring events at each unit time point in the specified historical time interval can be used as the time-varying λ of the service Poisson distribution. t , and use Gaussian process regression model to estimate λ t time-varying situation.

[0068] In one embodiment, Figure 2 As shown, step 130 may further include the following steps:

[0069] Step 130-1, input each unit time point within the specified historical time interval into the Gaussian process regression model, and obtain the instantaneous occurrence intensity output by the Gaussian process regression model for each unit time point, where the instantaneous occurrence intensity is the frequency of the monitoring event occurring at the current unit time point.

[0070] After the Gaussian process regression model is constructed, each unit time point within the specified historical time interval can be used as observation data to input into the Gaussian process regression model for prediction, and the instantaneous occurrence intensity corresponding to each unit time point can be obtained.

[0071] Step 130 - 2 , determining the average occurrence intensity of the monitored event at each unit time point within the specified historical time interval based on the historical occurrence data.

[0072] In this step, the average occurrence intensity of the monitored event at each unit time point in the specified historical time interval can be understood as the mean value of the parameter λ. In one implementation, the average occurrence intensity can be determined in the following manner: calculating a first number of unit time points in the specified historical time interval; calculating a second number of the monitored event occurring in the specified historical time interval; and taking the ratio of the first number to the second number as the average occurrence intensity.

[0073] For example, average occurrence intensity = t max / N, where t max is the first number of unit time points within the specified historical time interval, that is, the maximum value of the timestamp within the specified historical time interval; N is the second number of monitoring events occurring within the specified historical time interval.

[0074] Step 130 - 3 , for each unit time point, determine the occurrence frequency of the monitored event at the unit time point according to the instantaneous occurrence intensity at the current unit time point and the average occurrence intensity.

[0075] In this step, for each unit time point in the specified historical time interval, after obtaining the instantaneous occurrence intensity of each unit time point and the average occurrence intensity of the specified historical time interval, the two can be combined to determine the occurrence frequency of the current monitoring event in each unit time point.

[0076] In one implementation, step 130-3 may further include the following steps:

[0077] The instantaneous occurrence intensity at the current unit time point is mapped to a positive real number using a preset exponential function, and the product of the instantaneous occurrence intensity of the positive real number and the average occurrence intensity is calculated as the occurrence frequency of the monitored event at the unit time point.

[0078] In this step, since the Poisson process requires non-negative parameters, the output of the Gaussian process regression model (i.e., the instantaneous occurrence intensity) can be mapped to a positive real number using an exponential function in this embodiment. Calculating the product of the instantaneous occurrence intensity and the average occurrence intensity can make the instantaneous occurrence intensity λ learned by the model t Can fluctuate around λ.

[0079] Step 140 , generating a time-varying frequency curve for the monitoring event according to the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval, wherein the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time.

[0080] In this embodiment, the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time. After the occurrence frequency of each unit time point is obtained, the time-varying frequency curve of the current monitoring event can be fitted according to the occurrence frequency of each unit time point.

[0081] For example, Figure 3 As shown in the schematic diagram of the time-varying frequency curve, Figure 3 In , the x-axis is the normalized unit time point, and the y-axis is the occurrence frequency of each unit time point. According to the occurrence frequency of each unit time point, a smooth time-varying frequency curve can be fitted10. At the same time, for the convenience of comparison, Figure 3In the figure, a schematic diagram of the historical occurrence time of monitoring events is also drawn based on the historical occurrence time information 20. It can be seen that in the interval [0, 0.25]∪[0.75, 1.25]∪[1.75, 2], monitoring events occur more frequently, and in other time ranges, monitoring events occur more sparsely. Correspondingly, the learned λ t The variation curve also presents three peaks in the above three intervals and two valleys in the remaining intervals.

[0082] In one implementation, after obtaining the time-varying frequency curve of the current monitoring event, this embodiment may further include the following steps:

[0083] The time-varying frequency curve is sent to a downstream task, and the downstream task determines a high-frequency occurrence interval of a current monitoring event according to the time-varying frequency curve.

[0084] In this embodiment, the time-varying frequency curve can be sent to the downstream task, thereby providing assistance for the subsequent decision of the downstream task. A possible application example is to set a threshold t of λ h , when λ>t h , the corresponding interval can be defined as the high-frequency interval of the current monitoring event, so that when retrospective data analysis is performed, such a high-frequency interval can be automatically obtained, which is more efficient than manual identification. Figure 4 As shown, if t h = 20 as the dividing line, draw the intersection of the time-varying frequency curve and λ = 20, and the high-frequency interval can be automatically and quantitatively obtained. This interval is very close to the previously mentioned [0, 0.25] ∪ [0.75.1.25] ∪ [1.75, 2]. It can be imagined that when the amount of data is very large (tens of thousands or even hundreds of thousands), this method can very efficiently identify the high-frequency interval.

[0085] In order to enable those skilled in the art to better understand the embodiments of the present application, the following takes the monitoring event as a network connection fluctuation event (also known as a network failure event) as an example to illustrate the scenario of the present embodiment. Assume that the observed historical data is a series of standardized time points y1, y2, ... of network connection fluctuation events, and assume that the time intervals of these network connection fluctuation events are generated by sampling from a Poisson distribution. More specifically, it can be assumed that the parameter λ obeys the Poisson process t is not a fixed value, but a time-varying parameter. t The variation pattern of is unknown, so it is difficult to describe λ with an explicit function. t Based on this, in the embodiment of the present application, a non-parametric model, namely, a Gaussian process regression model, is used tot modeling the changing process.

[0086] Specifically, assume that the time point at which the i-th network connection fluctuation event occurs is y i (y i is the time information after standardization), and its generation process can be described as the following model:

[0087] y i -y i-1 ~Poisson(λ yi )

[0088] λ yi =g(y i )

[0089] g(y i )=exp[f(y i )]*(t max / N)

[0090] Among them, y i -y i-1 is the time interval between the occurrence of the ith network connection fluctuation event. The distribution of this time interval follows a time-varying Poisson distribution with parameter λ yi .

[0091] The second formula shows that λ yi is a time y i The function g(y i ). In this example, λ yi Characterizes the frequency of network connection fluctuation events, λ yi The higher the value, the higher the i Failures occur more frequently near this time point.

[0092] The third formula shows that g(y i ), where f(t) is a Gaussian process regression model. In this example, f(t) represents the frequency of network failures, so it can be called a failure intensity function. Since the Poisson process requires non-negative parameters, an exponential function is used in this example to map the output of the Gaussian process regression model to a positive real number. At the same time, multiply it by t max / N, where t max is the maximum value of the timestamp in the historical data (i.e., the first number of unit time points in the specified historical time interval), N is the number of failure events in the historical data, and t max / N is the average number of events per unit time in the historical data, that is, the average occurrence intensity, which is the mean value of parameter λ. Here, we multiply t max / N is the λ we want to learnt Can fluctuate around λ.

[0093] According to the description of the above model, If the ultimate goal is to find The core step of the expression is to find f(y i ), and in this embodiment, it is assumed that f can be described by a Gaussian process regression model, and f(y i ) is f in y i Sampling of points.

[0094] In this example, given In the case of , the likelihood function of the Gaussian process regression model f can be expressed as follows:

[0095]

[0096] Among them, y1,…,y n is the time variable observed at each n time, and T is the maximum value of the time variable in the observed data; is the λ at the time when the observed fault event data point occurs t The logarithm and is the expected number of failure events during the observation data period. The likelihood function can be understood as follows: when an event occurs, t The value of is high, but at the same time the average λ should be controlled t level, so that in the interval where events occur frequently λ t is high, and the interval λ where events occur sparsely t Low level. Minimize logp(y1, ..., y n |λ), we can get λ t The estimated value of .

[0097] exist Figure 3 In Figure 20, part 20 shows the time points of 51 network connection fluctuation events, which are normalized to [0, 2.0]. It can be seen that fluctuations occur more frequently in the interval [0, 0.25]∪[0.75, 1.25]∪[1.75, 2], and more sparsely in other time ranges. Figure 3 It can be seen from the curve of part 10 that the above model learns λ t The change shows 3 peaks and 2 valleys, and the interval where the peaks are located is very close to [0, 0.25] ∪ [0.75, 1.25] ∪ [1.75, 2] mentioned in Part 20.

[0098] When λ is obtained tAfter the change curve of , it can provide help for many downstream tasks and subsequent decisions. For example, a threshold t of λ can be given h , when λ>t h The corresponding interval can be defined as a high-fault or frequency interval, so that when performing retrospective data analysis, such a high-frequency interval can be automatically obtained, which is more efficient than manual identification.

[0099] In this embodiment, in the scenario of event monitoring, a Gaussian process regression model can be constructed based on the collected historical occurrence time information of the current monitoring event within the specified historical time interval, and then the occurrence frequency of the current monitoring event at each unit time point within the specified historical time interval can be determined based on the Gaussian process regression model, and a time-varying frequency curve for the monitoring event can be drawn based on the occurrence frequency. Based on the time-varying frequency curve, the change of the occurrence frequency of the current monitoring event over time can be determined, and then the high-frequency occurrence interval of the current monitoring event can be quickly identified, which is particularly suitable for scenarios with relatively large data volumes, and improves the recognition efficiency of monitoring events.

[0100] Embodiment 2

[0101] Figure 5 The structural block diagram of an event monitoring device embodiment provided in the second embodiment of the present application may include the following modules:

[0102] The historical data acquisition module 510 is used to acquire the historical occurrence data of the current monitoring event within a specified historical time interval, wherein the historical occurrence data includes the historical occurrence time information of the monitoring event;

[0103] A model building module 520, for building a Gaussian process regression model according to the historical occurrence time information;

[0104] An occurrence frequency determination module 530 is used to determine the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model, wherein the occurrence frequency obeys a Poisson distribution;

[0105] The frequency curve generating module 540 is used to generate a time-varying frequency curve for the monitoring event according to the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval, and the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time.

[0106] In one implementation, the occurrence frequency determination module 530 may further include the following submodules:

[0107] An instantaneous occurrence intensity determination submodule is used to input each unit time point within the specified historical time interval into the Gaussian process regression model, and obtain the instantaneous occurrence intensity output by the Gaussian process regression model for each unit time point, wherein the instantaneous occurrence intensity is the frequency of the monitoring event occurring at the current unit time point;

[0108] An average occurrence intensity determination submodule, used to determine the average occurrence intensity of the monitored event at each unit time point within the specified historical time interval according to the historical occurrence data;

[0109] The unit time point occurrence frequency determination submodule is used to determine the occurrence frequency of the monitored event at each unit time point according to the instantaneous occurrence intensity at the current unit time point and the average occurrence intensity.

[0110] In one embodiment, the average occurrence intensity determination submodule is specifically used for:

[0111] Calculating a first number of unit time points within the specified historical time interval;

[0112] Calculating a second number of occurrences of the monitored event within the specified historical time interval;

[0113] The ratio of the first quantity to the second quantity is taken as the average occurrence intensity.

[0114] In one implementation, the unit time point occurrence frequency determination submodule is specifically used to:

[0115] Using a preset exponential function to map the instantaneous occurrence intensity at the current unit time point into a positive real number;

[0116] The product of the instantaneous occurrence intensity of a positive real number and the average occurrence intensity is calculated as the occurrence frequency of the monitored event at the unit time point.

[0117] In one implementation, the historical occurrence time information is normalized time information.

[0118] In one implementation, the model building module 520 is specifically used to:

[0119] Taking the historical occurrence time information as a sample set, using the conjugate gradient method to minimize the log likelihood of the sample set, obtaining optimal hyperparameters, and determining a Gaussian process regression model according to the optimal hyperparameters;

[0120] According to the Gaussian process regression model, the corresponding posterior mathematical expectation function and posterior covariance function are determined to determine the posterior distribution probability.

[0121] In one embodiment, the device further comprises:

[0122] The curve sending module is used to send the time-varying frequency curve to the downstream task, and the downstream determines the high-frequency occurrence interval of the current monitoring event according to the time-varying frequency curve.

[0123] It should be noted that the above-mentioned event monitoring device provided in the embodiment of the present application can execute the event monitoring method provided in the first embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method.

[0124] Embodiment 3

[0125] Figure 6 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present application is shown in FIG. Figure 6 As shown, the electronic device includes a processor 610, a memory 620, an input device 630, and an output device 640; the number of the processor 610 in the electronic device can be one or more. Figure 6 A processor 610 is taken as an example; the processor 610, the memory 620, the input device 630 and the output device 640 in the electronic device can be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0126] The memory 620 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the method in the embodiment of the present application. The processor 610 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 620, that is, implementing the above method.

[0127] The memory 620 may mainly include a program storage area and a data storage area, wherein the program storage area

[0128] The operating system and the application required for at least one function may be stored; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 620 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 620 may further include a memory remotely arranged relative to the processor 610, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0129] The input device 630 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the electronic device. The output device 640 may include a display device such as a display screen.

[0130] Embodiment 4

[0131] Embodiment 4 of the present application further provides a storage medium comprising computer executable instructions, wherein the computer executable instructions are used to execute the method in any one of the embodiments in Embodiment 1 when executed by a processor of a server.

[0132] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present application can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0133] It is worth noting that in the embodiment of the above-mentioned device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application.

[0134] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. An event monitoring method, characterized in that: The method comprises: Obtaining historical occurrence data of a current monitoring event within a specified historical time interval, wherein the historical occurrence data includes historical occurrence time information of the monitoring event; Constructing a Gaussian process regression model based on the historical occurrence time information; Determine, according to the Gaussian process regression model, the frequency of occurrence of the monitoring event at each unit time point within the specified historical time interval, wherein the frequency of occurrence obeys a Poisson distribution; Generate a time-varying frequency curve for the monitoring event according to the frequency of occurrence of the monitoring event at each unit time point within the specified historical time interval, wherein the time-varying frequency curve is used to reflect the change of the frequency of occurrence of the monitoring event over time; Determining the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model includes: Input each unit time point within the specified historical time interval into the Gaussian process regression model, and obtain the instantaneous occurrence intensity output by the Gaussian process regression model for each unit time point, where the instantaneous occurrence intensity is the frequency of the monitoring event occurring at the current unit time point; Determine, based on the historical occurrence data, the average occurrence intensity of the monitored event at each unit time point within the specified historical time interval; For each unit time point, the occurrence frequency of the monitored event at the unit time point is determined according to the instantaneous occurrence intensity at the current unit time point and the average occurrence intensity.

2. The method according to claim 1, characterized in that Determining the average occurrence intensity of the monitored event at each unit time point within the specified historical time interval according to the historical occurrence data includes: Calculating a first number of unit time points within the specified historical time interval; Calculating a second number of occurrences of the monitored event within the specified historical time interval; The ratio of the first quantity to the second quantity is taken as the average occurrence intensity.

3. The method according to claim 1 or 2, characterized in that: The determining, for each unit time point, the occurrence frequency of the monitored event at the unit time point according to the instantaneous occurrence intensity at the current unit time point and the average occurrence intensity, includes: Using a preset exponential function to map the instantaneous occurrence intensity at the current unit time point into a positive real number; The product of the instantaneous occurrence intensity of a positive real number and the average occurrence intensity is calculated as the occurrence frequency of the monitored event at the unit time point.

4. The method according to claim 1, characterized in that: The historical occurrence time information is time information that has been normalized.

5. The method according to claim 1, 2 or 4, characterized in that: The constructing of a Gaussian process regression model according to the historical occurrence time information includes: Taking the historical occurrence time information as a sample set, using the conjugate gradient method to minimize the log likelihood of the sample set, obtaining optimal hyperparameters, and determining a Gaussian process regression model according to the optimal hyperparameters; According to the Gaussian process regression model, the corresponding posterior mathematical expectation function and posterior covariance function are determined to determine the posterior distribution probability.

6. The method according to claim 1, characterized in that The method further comprises: The time-varying frequency curve is sent to a downstream task, and the downstream task determines a high-frequency occurrence interval of a current monitoring event according to the time-varying frequency curve.

7. An event monitoring device, characterized in that: The device comprises: A historical data acquisition module is used to acquire historical occurrence data of a current monitoring event within a specified historical time interval, wherein the historical occurrence data includes historical occurrence time information of the monitoring event; A model building module, used to build a Gaussian process regression model according to the historical occurrence time information; An occurrence frequency determination module, used to determine the occurrence frequency of the monitoring event at each unit time point within the specified historical time interval according to the Gaussian process regression model, wherein the occurrence frequency obeys a Poisson distribution; A frequency curve generating module, used to generate a time-varying frequency curve for the monitoring event according to the frequency of occurrence of the monitoring event at each unit time point within the specified historical time interval, wherein the time-varying frequency curve is used to reflect the change of the occurrence frequency of the monitoring event over time; The occurrence frequency determination module includes an instantaneous occurrence intensity determination submodule, an average occurrence intensity determination submodule and a unit time point occurrence frequency determination submodule; The instantaneous occurrence intensity determination submodule is used to input each unit time point within the specified historical time interval into the Gaussian process regression model, and obtain the instantaneous occurrence intensity output by the Gaussian process regression model for each unit time point, wherein the instantaneous occurrence intensity is the frequency of the monitoring event occurring at the current unit time point; The average occurrence intensity determination submodule is used to determine the average occurrence intensity of the monitored event at each unit time point within the specified historical time interval according to the historical occurrence data; The unit time point occurrence frequency determination submodule is used to determine, for each unit time point, the occurrence frequency of the monitored event at the unit time point according to the instantaneous occurrence intensity at the current unit time point and the average occurrence intensity.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.