Private communication anomaly detection method and system

By using long and short-term memory network algorithms in the privacy communication system to build a detection index prediction model and analyze call data in real time, the problems of slow response speed and low detection efficiency in the existing technology are solved, and more accurate and timely abnormal detection is achieved, improving System stability and operation and maintenance efficiency.

CN119031070APending Publication Date: 2024-11-26CHINA UNICOM ONLINE INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411324846.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-23
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing privacy communication system responds slowly after an exception occurs, making it difficult to process large-scale and high-dimensional call data, resulting in limited abnormal detection efficiency and accuracy.

Method used

A long-term and short-term memory network algorithm is used to build a detection index prediction model, calculate dynamic thresholds in real time, and automatically identify potential faults by collecting and analyzing call data in real time, triggering early warning mechanisms in a timely manner, reducing the need for manual intervention.

Benefits of technology

It improves the timeliness and accuracy of abnormal detection, reduces the impact time of failure on the system, reduces operation and maintenance costs, and improves the stability and response efficiency of privacy account services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119031070A_ABST
    Figure CN119031070A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of privacy communication, and provides a privacy communication anomaly detection method and system. The method comprises the steps that call ticket data are collected from a communication service providing platform and preprocessed, and the call ticket data comprise the total call volume of privacy numbers; constructing a detection index prediction model by using a long short-term memory network algorithm, calculating a dynamic threshold value in real time, and making result judgment based on the dynamic threshold value; determining model input data according to a to-be-judged time period, and inputting the model input data into the detection index prediction model to obtain an index prediction value corresponding to the to-be-judged time period; the actual detection index value is called, and detection index abnormity judgment is carried out according to the actual detection index value, the index prediction value and the dynamic threshold value calculated in real time; and determining a trigger early warning mechanism according to a detection index abnormity judgment result. According to the invention, real-time automatic anomaly detection and early warning can be realized, the requirement of manual intervention is reduced, the operation and maintenance cost is reduced, and the overall operation and maintenance efficiency of the system is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of privacy communication technology, and in particular to a privacy communication anomaly detection method and system. Background Art

[0002] Privacy communication is to meet the rigid demands of customers such as "privacy protection, enterprise data security, and service quality control". It protects the real number information of both parties by providing intermediate numbers and establishing communication between the caller and the called party. However, in the process of providing communication services, the following problems are often encountered: the provincial node that provides privacy communication service capabilities suddenly fails, resulting in all or part of the numbers under the provincial node being unable to be used normally, which will lead to customer complaints and affect customer business development.

[0003] With the rapid development of artificial intelligence technology, neural network algorithms have demonstrated powerful capabilities in the field of data processing and analysis. For anomaly detection on privacy communication platforms, by real-time analysis of changing trends in call data and historical comparisons, they can automatically identify and predict potential faults, achieve rapid and accurate judgment of system status, and improve system stability and reliability.

[0004] The shortcomings of the current method for discovering system anomalies and provincial node anomalies are: after the customer finds that the number cannot be dialed during business use, the customer manually reports the anomaly; the operator checks the connection rate of the key customers' numbers, and screens the provincial nodes that provide privacy communication services for a large number of numbers with a connection rate of 0, thereby discovering the anomaly. The number and time of privacy communication service calls and the number of numbers in the provincial nodes are strongly correlated. The operator can manually determine whether the privacy communication platform and provincial nodes are abnormal based on the time period and the call volume of different provincial nodes. However, the existing processing method can only trigger an alarm after a period of time after the anomaly occurs, and the overall response speed is slow, which also affects the timeliness of fault handling. In addition, it is difficult to process large-scale, high-dimensional call data, which limits the efficiency and accuracy of anomaly detection.

[0005] Therefore, it is necessary to provide a privacy communication anomaly detection method and system to solve the above problems. Summary of the invention

[0006] The present invention aims to provide a privacy communication anomaly detection method and system to solve the technical problems in the prior art that the existing methods can only trigger an alarm after a period of time after the anomaly occurs, the overall response speed is slow, and it also affects the timeliness of fault handling. It is difficult to process large-scale, high-dimensional call data, which leads to limited efficiency and accuracy of anomaly detection. The technical problem to be solved by the present invention is achieved through the following technical solution.

[0007] The first aspect of the present invention proposes a privacy communication anomaly detection method, comprising: collecting call bill data from a communication service provider platform, and preprocessing the call bill data, the call bill data including the total call volume of the privacy number, the preprocessing including constructing a feature vector based on the timestamp; using a long short-term memory network algorithm to construct a detection index prediction model, and calculating a dynamic threshold in real time, and making a result judgment based on the dynamic threshold, the detection index including the total call volume of the privacy number; determining the model input data according to the time period to be judged, inputting the determined model input data into the detection index prediction model, and obtaining an index prediction value corresponding to the time period to be judged; retrieving the actual detection index value of the time period to be judged, and making a detection index anomaly judgment based on the actual detection index value, the index prediction value and the dynamic threshold calculated in real time; determining the triggering early warning mechanism according to the result of the detection index anomaly judgment.

[0008] The second aspect of the present invention proposes a privacy communication anomaly detection system, which adopts the privacy communication anomaly detection method described in the first aspect of the present invention, and the privacy communication anomaly detection system includes: a data collection module, which is used to collect call list data from a communication service provider platform and pre-process the call list data, wherein the call list data includes the total call volume of the privacy number, and the pre-processing includes constructing a feature vector according to the timestamp; a model construction module, which is used to use a long short-term memory network algorithm to construct a detection index prediction model for predicting the detection index, and calculate the dynamic threshold in real time, and make a result judgment based on the dynamic threshold, wherein the detection index includes the total call volume of the privacy number; a prediction module, which determines the model input data according to the time period to be judged, inputs the determined model input data into the detection index prediction model, and obtains the index prediction value corresponding to the time period to be judged; a judgment module, which is used to retrieve the actual detection index value of the time period to be judged, and makes a detection index abnormality judgment according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time; an early warning module, which determines the triggering of the early warning mechanism according to the result of the detection index abnormality judgment.

[0009] The third aspect of the present invention provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the privacy communication anomaly detection method described in the first aspect of the present invention.

[0010] A fourth aspect of the present invention provides a computer-readable medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for detecting anomalies in private communications described in the first aspect of the present invention is implemented.

[0011] The embodiments of the present invention include the following advantages: Compared with the prior art, the present invention can identify and warn at the early stage of anomalies by real-time collection and analysis of call data, thereby reducing the impact time of faults on the system and improving the timeliness of anomaly detection; a neural network algorithm is used to utilize historical call list data (specifically including the total call volume); a long short-term memory network algorithm is used to construct a detection index prediction model, so that the model can automatically learn to identify the characteristics of whether the communication service provider platform or sub-platform is operating normally and in an abnormal state, thereby improving the accuracy of anomaly detection; based on the actual detection index value, the index prediction value and the dynamic threshold calculated in real time, the detection index abnormality judgment is made, which can more accurately determine whether the communication service provider platform or sub-platform is abnormal (i.e., in an abnormal state), can trigger the early warning mechanism more timely, and notify maintenance personnel to troubleshoot and handle the fault, which can reduce the impact on users and the cost of manual troubleshooting, can improve the efficiency of abnormal response, and can also improve the stability of the privacy number service.

[0012] In addition, through real-time automated anomaly detection and early warning, the need for manual intervention is reduced, operation and maintenance costs are reduced, and the overall operation and maintenance efficiency of the system is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a flowchart of an example of a method for detecting anomalies in private communications according to the present invention; Figure 2 is a schematic diagram of an application example of the privacy communication anomaly detection method of the present invention; Figure 3 is a structural block diagram of the privacy communication anomaly detection system of the present invention; Figure 4 is a schematic structural diagram of an electronic device embodiment according to the present invention; Figure 5 It is a schematic diagram of the structure of a computer-readable medium embodiment according to the present invention. DETAILED DESCRIPTION

[0014] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0015] In view of the above problems, the present invention proposes a privacy communication anomaly detection method, which can identify and warn at the early stage of the occurrence of anomalies by real-time collection and analysis of call data, which can reduce the impact time of faults on the system, and thus improve the timeliness of anomaly detection; adopt a neural network algorithm to use historical call data (specifically including the total call volume); use a long short-term memory network algorithm to construct a detection index prediction model, which can obtain more accurate index prediction values, and construct a detection index anomaly judgment model, so that the model can automatically learn to identify whether the communication service provider platform or sub-platform is operating normally and in an abnormal state, so that it can accurately identify whether the detection index is abnormal, thereby improving the accuracy of anomaly detection; according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time, the detection index anomaly judgment is performed, which can more accurately determine whether the communication service provider platform or sub-platform is abnormal (i.e., abnormal state), and can trigger the early warning mechanism more timely to notify maintenance personnel to troubleshoot and handle the fault, which can reduce the impact on users and the cost of manual troubleshooting, improve the efficiency of abnormal response, and improve the stability of privacy number services.

[0016] Example 1 Refer to the following Figure 1 and Figure 2 , the contents of the present invention will be described in detail.

[0017] Figure 1 It is a flowchart of an example of the privacy communication anomaly detection method of the present invention.

[0018] like Figure 1 As shown, in step S101, call bill data is collected from a communication service provider platform and pre-processed, wherein the call bill data includes the total call volume of a private number, and the pre-processing includes constructing a feature vector based on a timestamp.

[0019] In a specific application example, Figure 2 As shown, it includes a communication service provider platform and multiple sub-platforms (for example, five sub-platforms) that can communicate with the communication service provider platform in real time. The communication service provider platform and multiple sub-platforms can provide privacy numbers (i.e., intermediate numbers) and related privacy communication services for users such as food delivery, online car-hailing, express delivery, and financial services, so that the caller and the called party can establish communication and protect the real numbers of both parties while using the privacy numbers normally based on the communication service provider platform or sub-platform.

[0020] Specifically, the time series data is processed according to the time window slicing method. Optionally, the call data of the communication service providing platform or sub-platform is collected according to a specified frequency, and the specified frequency includes ten minutes, half an hour, one hour, 12 hours, and one day, which is determined according to the user or the communication service business.

[0021] In a specific implementation, the call record data of the communication service provider platform or sub-platform is collected every ten minutes to collect the total call volume, the total call volume of each sub-platform (for example, each sub-platform corresponds to a provincial node), the timestamp, and the code of each sub-platform.

[0022] It should be noted that call volume data is greatly affected by time, region and business type. For example, the call volume is higher during the day and lower at night. The call volume of takeaway business is higher during the three meal periods every day, and the call volume of taxi business is higher during the morning and evening peak hours. Therefore, the precision tendency selects the corresponding data for processing, which will be used for subsequent model training data or other subsequent data processing.

[0023] Next, the call record data is preprocessed, specifically including timestamp processing and total call volume processing.

[0024] For timestamp processing, the timestamp is converted into a comparable date and time format, and the hour and minute information is extracted to match the time period.

[0025] For call volume processing, the total call volume of the communication service provider platform (or sub-platform) collected is cleaned and standardized to eliminate noise and outliers. For missing values, since the total call volume is highly correlated with time, the mean filling method can be used. For example, for data missing at a certain time scale, the mean of the call volume in the same time period within seven days is inserted to fill it. The call volume time series data is Gaussian transformed to form a feature vector (i.e., construct a feature vector) to be more suitable for learning features when training the model.

[0026] For example, the total call volume of the communication service provider platform and the total call volume of the sub-platform (i.e., the nodes in each province) are collected at a frequency of once every ten minutes, and the data of the past two months are collected as the historical data set D, D = [T, N S ,N px ,x=1,2,3,…,n,P], where T is the timestamp set, N S The number of platform calls provided for communication services, N px is the total number of calls of a sub-platform deployed in a certain province, and P is the sub-platform code, which is used to distinguish the call volume of different provinces in subsequent analysis.

[0027] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0028] Next, in step S102, a detection index prediction model is constructed using a long short-term memory network algorithm, and a dynamic threshold is calculated in real time, and a result judgment is made based on the dynamic threshold. The detection index includes the total call volume of the privacy number.

[0029] Specifically, a detection index prediction model is constructed using the long short-term memory network (LSTM) algorithm to predict the detection index.

[0030] The detection index prediction model includes an input layer, a hidden layer and an output layer. Since the prediction task of the detection index involves fewer parameters, the hidden layer is set to 2 to enhance the nonlinear fitting ability of the model and accelerate the convergence speed of the model. However, it is not limited to this. In other prediction tasks, the number of hidden layers can also be adaptively adjusted. The above is only described as a preferred example and cannot be understood as a limitation of the present invention.

[0031] It should be noted that the detection index prediction model constructed using the long short-term memory network algorithm can better capture and utilize the long-term dependencies in time series data. Therefore, the long short-term memory network algorithm is selected to construct the detection index prediction model, which can be simplified into a function call to represent the detection index prediction model: =LSTM_Model(T i , N pxi , P i ); in, is the total call volume of the communication service provider platform or sub-platform output by the detection indicator prediction model, T i Represents a timestamp sequence collected at a specified frequency within a specified historical period. LSTM_Model refers to an indicator detection model built using the LSTM algorithm. N pxi The total call volume sequence collected by the communication service provider platform or each sub-platform within a specified historical time period and at a specified frequency (for example, a total call volume sequence of 18 in length collected every ten minutes within three hours); i The encoding for a sub-platform.

[0032] In the first embodiment, appropriate time series data is selected according to the historical data set obtained in step S101 and used as model input data in the form of multi-dimensional input data.

[0033] According to the time period to be determined, the following data are determined to form model input data: a timestamp sequence, a historical detection indicator sequence, and a coding sequence of the communication service provider platform or a sub-platform coding sequence.

[0034] A training data set is established to train a detection index prediction model using a timestamp sequence marked with a call volume label, a historical detection index quantity sequence, and a coding sequence of the communication service provider platform or sub-platform. The call volume label specifically refers to the total call volume corresponding to the timestamp sequence.

[0035] Specifically, the historical data set D is divided into time windows of a certain length to generate a training data set D train Assuming that the time window length is 18 call volume data at a time, that is, the total call volume (that is, the data collected within 3 hours), each training data is generated as follows: , where T i N is a sequence of timestamps every ten minutes within three hours; pxi A total call volume sequence of length 18 collected every ten minutes within three hours for the communication service provider platform or each sub-platform; T i The actual total call volume of the next collection point; P i Provides a platform code sequence or each sub-platform code sequence for a communication service of length 18.

[0036] It should be noted that in this example, the training data set is established with the above sequence data marked with the total call volume sequence of the communication service provider platform or each sub-platform in a specified historical time period (for example, the length of 18 collected every ten minutes within three hours), specifically including a timestamp sequence (including the historical time period, collection frequency, etc.), a communication service provider platform code sequence or each sub-platform code sequence. The above is only explained as an optional example and cannot be understood as a limitation of the present invention.

[0037] The following loss function is used to optimize the training process of the detection index prediction model: ; in, represents the difference between the total call volume prediction value output by the detection index prediction model and the actual total call volume, i.e., the loss value; Y represents the actual total call volume; f(X) represents the total call volume prediction value output by the detection index prediction model, and X is the model input data of the detection index prediction model.

[0038] When the difference is less than 1% (i.e., the limiting condition for stopping training), the training of the detection index prediction model is stopped, and the trained detection index prediction model is obtained. Therefore, a more accurate detection index prediction model can be obtained.

[0039] It should be noted that, in other implementations, techniques such as regularization and dropout algorithms may be introduced to prevent overfitting and improve the generalization ability of the model.

[0040] The second embodiment is different from the first embodiment in that a detection index abnormality judgment model is constructed to calculate a dynamic threshold in real time, and a result judgment is made based on the dynamic threshold.

[0041] A deep learning neural network algorithm is used to build a detection indicator abnormality judgment model.

[0042] The detection index abnormality judgment model is used to learn whether the ratio of the difference between the total call volume predicted by the model and the actual total call volume obtained is within the set threshold range, that is, directly output whether the total call volume of the communication service provider platform or sub-platform is abnormal. In other words, a mapping relationship is established between whether the total call volume of the communication service provider platform or sub-platform is abnormal and the time period to be judged, so that when the model is applied for prediction, the time period to be judged is input, and whether the total call volume of the communication service provider platform or sub-platform is abnormal is output.

[0043] Specifically, a timestamp sequence marked with normal indicator labels and abnormal indicator labels, a historical detection indicator quantity sequence, and a coding sequence of the communication service provider platform or sub-platform are used to establish a training data set and train a detection indicator abnormality judgment model. The normal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is within the set threshold range, and the abnormal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is not within the set threshold range.

[0044] The loss function for optimizing the training process of the detection index abnormality judgment model is the same as the loss function for model training in the first embodiment. Alternatively, whether the ratio of the number of correct abnormality judgments of the indicator to the total number of judgments is greater than a specified value, for example, the specified value is 0.98 to 0.99. The limiting condition for stopping training is specifically that when the accuracy of abnormality judgment of the detection index abnormality judgment model is greater than 99%, the training model is stopped. Therefore, the description of the same part is omitted.

[0045] Next, in step S103, model input data is determined according to the time period to be determined, and the determined model input data is input into the detection index prediction model to obtain the index prediction value corresponding to the time period to be determined.

[0046] In a first embodiment, model input data is determined according to the time period to be determined, and the determined model input data is input into the detection index prediction model to obtain an index prediction value corresponding to the time period to be determined.

[0047] For example, if the time period to be judged is from 11:00 a.m. to 11:30 a.m., determine the timestamp sequence (specifically, count forward two hours from 11:00 a.m. and the stimulation frequency is once every half an hour), obtain the historical total call volume sequence corresponding to the timestamp sequence, and obtain the communication service provider platform coding sequence or sub-platform coding sequence to form the model input data.

[0048] The model input data formed above is input into the detection index prediction model to obtain the index prediction value corresponding to the time period to be determined.

[0049] Next, the dynamic threshold of the detection index in each time period is dynamically calculated using the historical communication service business data. The dynamic threshold is used to determine whether the detection index is abnormal. Specifically, the dynamic threshold of the detection index is calculated using the following expression: =Mean(Historical_Calls,time−14days≤time′≤time−1day) in, Indicates the average of the dynamic thresholds of the calculated detection indicators in multiple specified historical time periods, the detection indicators include the call volume of the communication service provider platform in the specified historical time period, and the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period. The specified historical time period includes the time period that is the same as time in each day within 14 days from the current time. The specified historical time period is a periodic time period; Historical_Calls indicates the number of dynamic thresholds of the communication service provider platform every day with the historical timestamp The call volume data set with historical timestamps in the same time period, or the call volume data of each sub-platform every day The call volume data set whose historical timestamps are in the same time period; N pxi is the total call volume sequence of the communication service provider platform or each sub-platform at a specified frequency, i.e., the collection frequency, within a specified historical period; time′ represents the total call volume sequence of the communication service provider platform every day. The historical timestamp is the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day, or the sampling time of each sub-platform every day. The historical timestamp is located at the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day.

[0050] In an optional implementation, detection indicators of the communication service provider platform or multiple sub-platforms in multiple historical time periods in historical communication service business data are used to establish a first segmented curve trend graph and a second segmented curve trend graph corresponding to the communication service provider platform and each sub-platform respectively.

[0051] Specifically, around midnight, fewer people generally use privacy numbers, and the total call volume trend tends to be a straight line. During the day, due to different business classifications and different times of using privacy numbers, the total call volume will show different trends (including straight lines, curves, such as multiple curves and multiple straight lines, straight lines between multiple curves, etc.). According to different user types, different privacy number types, time periods, and the total call volume of privacy numbers in each time period, a first segmented curve trend graph corresponding to the communication service provider platform is drawn, and a second segmented curve trend graph corresponding to each sub-platform is drawn.

[0052] The dynamic threshold is calculated based on the first segmented curve trend graph or the second segmented curve trend graph formed by different trends in each time period of each day, which improves the accuracy of the dynamic threshold calculation and the accuracy of the final abnormal result judgment. The detection index is, for example, the total call volume.

[0053] According to the obtained first segmented curve trend graph and the second segmented curve trend graph, the dynamic threshold of the detection index in each time period is calculated in real time.

[0054] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0055] Next, in step S104, the actual detection index value of the time period to be judged is retrieved, and the detection index abnormality judgment is performed based on the actual detection index value, the index prediction value and the dynamic threshold calculated in real time.

[0056] Specifically, the detection index abnormality judgment is performed based on the actual detection index value, the index prediction value and the dynamic threshold calculated in real time.

[0057] Further, the difference between the actual detected index value and the predicted value of the index is calculated.

[0058] Specifically, the following expression is used to calculate the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform at the current time to the dynamic threshold: : ; in, Indicates the ratio of the predicted deviation of the total call volume of the communication service provider platform or sub-platform at the current time to the dynamic threshold; Indicates the average of the dynamic thresholds of the calculated detection indicators in multiple specified historical time periods, the detection indicators include the call volume of the communication service provider platform in the specified historical time period, and the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period, the specified historical time period includes the time period that is the same as time in each day within 14 days from the current time, and the specified historical time period is a periodic time period; The predicted total call volume data of the collection point corresponding to the current time; It is the actual total call volume data of the collection point corresponding to the current time.

[0059] In an optional implementation, when the ratio of the calculated total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold If it is not within the set threshold range, it is determined that the communication service provision capability of the communication service provision platform at the current time time is abnormal or the communication service provision capability of the sub-platform is abnormal.

[0060] When the ratio of the calculated total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold When the threshold range is set, it is determined that the communication service provision capability of the communication service provision platform at the current time is normal or the communication service provision capability of the sub-platform is normal.

[0061] In another optional implementation, when the ratio of the calculated total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold When it is not within the set threshold range, and when the number of times it is not within the set threshold range within a specific time interval (for example, when the sampling frequency is 10 minutes, two hours is used as the time interval for number determination) is greater than the number threshold, it is determined that the communication service provision capability of the communication service provision platform or sub-platform is abnormal.

[0062] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0063] Next, in step S105, the triggering of the early warning mechanism is determined according to the abnormality judgment result of the detection index.

[0064] According to the abnormality judgment result of the detection indicator, specifically when the detection indicator is abnormal, it is determined that the communication service providing platform or sub-platform is abnormal, and an early warning mechanism is triggered. The early warning mechanism includes fault handling and number migration processing.

[0065] For example, if multiple judgment results are judged to be abnormal within a certain time interval, for example, the number of abnormal judgments exceeds a certain proportion (for example, 30%) within two hours, then the communication service capability of the sub-platform is determined to be abnormal, and the early warning mechanism is immediately triggered, such as sending an alarm text message to the operator and the person in charge of the sub-platform to notify the relevant personnel to handle it and quickly migrate the number (for example, transfer all private numbers on the sub-platform to an available sub-platform) to restore business as soon as possible, thereby reducing user losses.

[0066] Optionally, the process and results of fault detection are recorded for subsequent analysis and model optimization.

[0067] In addition, the real-time judgment results are statistically compiled into a visual line graph and intuitively displayed on the relevant visualization page. Operators can log in to the privacy communication platform and clearly see the analysis results of the communication service capabilities of each sub-platform on the relevant visualization page.

[0068] Through the above process, the communication service capabilities of the communication service provider platform or sub-platform can be tested in real time, and abnormalities or failures that occur can be quickly perceived and handled, thereby improving user service quality and reducing labor costs.

[0069] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0070] In addition, the drawings are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the drawings do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0071] Compared with the prior art, the present invention can identify and warn at the early stage of anomaly occurrence by real-time collection and analysis of call data, which can reduce the impact time of faults on the system, and thus improve the timeliness of anomaly detection; a neural network algorithm is used to utilize historical call list data (specifically including the total call volume); a long short-term memory network algorithm is used to construct a detection index prediction model, which can obtain more accurate index prediction values, and a detection index abnormality judgment model is constructed, so that the model can automatically learn to identify the characteristics of whether the communication service provider platform or sub-platform is operating normally and in an abnormal state, so that it can accurately identify whether the detection index is abnormal, thereby improving the accuracy of anomaly detection; according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time, the detection index abnormality judgment is performed, which can more accurately determine whether the communication service provider platform or sub-platform is abnormal (i.e., abnormal state), and can trigger the early warning mechanism more timely to notify maintenance personnel to troubleshoot and handle the fault, which can reduce the impact on users and the cost of manual troubleshooting, improve the efficiency of abnormal response, and improve the stability of privacy number services.

[0072] In addition, through real-time automated anomaly detection and early warning, the need for manual intervention is reduced, operation and maintenance costs are reduced, and the overall operation and maintenance efficiency of the system is improved.

[0073] Example 2 The following is a system embodiment of the present invention, which can be used to implement the method embodiment of the present invention. For details not disclosed in the system embodiment of the present invention, please refer to the method embodiment of the present invention.

[0074] Figure 3 It is a structural diagram of an example of a privacy communication anomaly detection system.

[0075] Reference Figure 3 The second aspect of the present disclosure provides a privacy communication anomaly detection system 300, which adopts the privacy communication anomaly detection method described in the first aspect of the present invention.

[0076] Specifically, the privacy communication anomaly detection system 300 includes a data collection module 310, a model building module 320, a prediction module 330, a judgment module 340 and an early warning module 350.

[0077] In a specific embodiment, the data collection module 310 is used to collect call data from the communication service provider platform and pre-process the call data, wherein the call data includes the total call volume of the privacy number, and the pre-processing includes constructing a feature vector according to the timestamp. The model construction module 320 is used to construct a detection index prediction model using a long short-term memory network algorithm, which is used to predict the detection index, wherein the detection index includes the total call volume of the privacy number, and judges the abnormal results in real time based on the dynamic threshold. The prediction module 330 determines the model input data according to the time period to be judged, inputs the determined model input data into the detection index prediction model, and obtains the index prediction value corresponding to the time period to be judged. The judgment module 340 is used to retrieve the actual detection index value of the time period to be judged, and judges the abnormality of the detection index according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time. The early warning module 350 determines the triggering of the early warning mechanism according to the abnormal judgment result of the detection index.

[0078] In a specific implementation, the abnormality judgment of the detection index according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time includes: The following expression is used to calculate the ratio of the predicted deviation of the total call volume of the communication service provider platform or sub-platform at the current time to the dynamic threshold: : ; in, Indicates the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform at the current time to the dynamic threshold. When calculating the total call volume prediction deviation of the communication service provider platform, p is used to represent the communication service provider platform; when calculating the total call volume of the sub-platform, p is used to represent the sub-platform; Indicates the average of the dynamic thresholds of the calculated detection indicators in multiple specified historical time periods, the detection indicators include the call volume of the communication service provider platform in the specified historical time period, and the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period, the specified historical time period includes the time period that is the same as the current time in each day within 14 days from the current time, and the specified historical time period is a periodic time period; The predicted total call volume data of the collection point corresponding to the current time; It is the actual total call volume data of the collection point corresponding to the current time.

[0079] When the calculated ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold If it is not within the set threshold range, it is determined that the communication service provision capability of the communication service provision platform at the current time time is abnormal or the communication service provision capability of the sub-platform is abnormal.

[0080] When the calculated ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold When the threshold range is set, it is determined that the communication service provision capability of the communication service provision platform at the current time is normal or the communication service provision capability of the sub-platform is normal.

[0081] According to an optional implementation, the detection index abnormality judgment is performed based on the actual detection index value, the index prediction value and the dynamic threshold calculated in real time.

[0082] Specifically, when the ratio of the calculated total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold If it is not within the set threshold range, and when the number of times it is not within the set threshold range within a specific time interval is greater than the number threshold, it is determined that the communication service provision capability of the communication service provision platform or sub-platform is abnormal.

[0083] According to an optional implementation manner, historical communication service business data is used to dynamically calculate the dynamic threshold of the detection index in each time period, and the dynamic threshold is used to determine whether the detection index is abnormal; specifically, the dynamic threshold of the detection index is calculated using the following expression: =Mean(Historical_Calls,time−14days≤time′≤time−1day) in, Indicates the average of the dynamic thresholds of the calculated detection indicators in multiple specified historical time periods, the detection indicators include the call volume of the communication service provider platform in the specified historical time period, and the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period, the specified historical time period includes the time period that is the same as time in each day within 14 days from the current time, and the specified historical time period is a periodic time period; Historical_Calls indicates the number of calls that the communication service provider platform receives from the communication service provider platform every day. The call volume data set with the historical timestamp in the same time period, or the call volume data of each sub-platform every day A call volume data set whose historical timestamps are in the same time period; is the total call volume sequence of the communication service provider platform or each sub-platform at a specified frequency, i.e., the collection frequency, within a specified historical period; time′ represents the total call volume sequence of the communication service provider platform every day. The historical timestamp is the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day, or the sampling time of each sub-platform every day. The historical timestamp is located at the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day.

[0084] According to an optional implementation manner, detection indicators of the communication service providing platform or multiple sub-platforms in multiple historical time periods in historical communication service business data are used to establish a first segmented curve trend graph and a second segmented curve trend graph corresponding to the communication service providing platform and each sub-platform respectively; the detection indicators include the total call volume.

[0085] According to the obtained first segmented curve trend graph and the second segmented curve trend graph, the dynamic threshold of the detection index in each time period is calculated in real time.

[0086] According to an optional implementation, model input data is determined according to the time period to be determined, and the determined model input data is input into the detection index prediction model to obtain an index prediction value corresponding to the time period to be determined.

[0087] According to the time period to be judged, the following data are determined to form model input data: a timestamp sequence, a historical detection indicator sequence, and a coding sequence of the communication service providing platform or sub-platform.

[0088] A training data set is established to train a detection index prediction model using a timestamp sequence marked with a call volume label, a historical detection index quantity sequence, and a coding sequence of the communication service provider platform or sub-platform. The call volume label specifically refers to the total call volume corresponding to the timestamp sequence.

[0089] According to an optional implementation manner, a detection indicator abnormality judgment model is established, and a timestamp sequence marked with normal indicator labels and abnormal indicator labels, a historical detection indicator quantity sequence, and a coding sequence of the communication service provider platform or sub-platform are used to establish a training data set and train the detection indicator judgment model. The normal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is within the set threshold range, and the abnormal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is not within the set threshold range.

[0090] According to an optional implementation manner, the following loss function is used to optimize the training process of the detection index prediction model: ; in, represents the difference between the total call volume prediction value output by the detection index prediction model and the actual total call volume, i.e., the loss value; Y represents the actual total call volume; f(X) represents the total call volume prediction value output by the detection index prediction model, and X is the model input data of the detection index prediction model.

[0091] When the difference is less than 1%, the training of the detection index prediction model is stopped, and the trained detection index prediction model is obtained.

[0092] According to an optional implementation manner, based on the abnormality judgment result of the detection indicator, specifically when the detection indicator is abnormal, it is determined that the communication service providing platform or sub-platform is abnormal, and an early warning mechanism is triggered. The early warning mechanism includes fault handling and number migration processing.

[0093] Compared with the prior art, the present invention can identify and warn at the early stage of anomaly occurrence by real-time collection and analysis of call data, which can reduce the impact time of faults on the system, and thus improve the timeliness of anomaly detection; a neural network algorithm is used to utilize historical call list data (specifically including the total call volume); a long short-term memory network algorithm is used to construct a detection index prediction model, which can obtain more accurate index prediction values, and a detection index abnormality judgment model is constructed, so that the model can automatically learn to identify the characteristics of whether the communication service provider platform or sub-platform is operating normally and in an abnormal state, so that it can accurately identify whether the detection index is abnormal, thereby improving the accuracy of anomaly detection; according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time, the detection index abnormality judgment is performed, which can more accurately determine whether the communication service provider platform or sub-platform is abnormal (i.e., abnormal state), and can trigger the early warning mechanism more timely to notify maintenance personnel to troubleshoot and handle the fault, which can reduce the impact on users and the cost of manual troubleshooting, improve the efficiency of abnormal response, and improve the stability of privacy number services.

[0094] In addition, through real-time automated anomaly detection and early warning, the need for manual intervention is reduced, operation and maintenance costs are reduced, and the overall operation and maintenance efficiency of the system is improved.

[0095] It should be noted that the privacy communication anomaly detection method performed by the privacy communication anomaly detection in Example 2 is the same as the privacy communication anomaly detection method in Example 1, so the description of the same parts is omitted.

[0096] Figure 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

[0097] like Figure 4As shown, the electronic device is presented in the form of a general computing device. The processor may be one or more and work in coordination. The present invention does not exclude distributed processing, that is, the processor may be dispersed in different physical devices. The electronic device of the present invention is not limited to a single entity, but may also be the sum of multiple physical devices.

[0098] The memory stores a computer executable program, which is usually a machine-readable code. The computer-readable program can be executed by the processor to enable the electronic device to perform the method of the present invention, or at least part of the steps in the method.

[0099] The memory includes a volatile memory, such as a random access memory unit (RAM) and / or a cache memory unit, and may also be a non-volatile memory, such as a read-only memory unit (ROM).

[0100] Optionally, in this embodiment, the electronic device further includes an I / O interface, which is used for the electronic device to exchange data with an external device. The I / O interface can represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of a variety of bus structures.

[0101] It should be understood that Figure 4 The electronic device shown is only an example of the present invention, and the electronic device of the present invention may also include elements or components not shown in the above examples. For example, some electronic devices also include display units such as display screens, and some electronic devices also include human-computer interaction elements such as buttons, keyboards, etc. As long as the electronic device can execute the computer-readable program in the memory to implement the method of the present invention or at least part of the steps of the method, it can be considered as an electronic device covered by the present invention.

[0102] Through the above description of the implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by combining software with necessary hardware. Figure 5 As shown, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of commands to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the embodiment of the present invention.

[0103] The software product may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0104] The computer readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with a command execution system, device, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0105] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0106] The computer-readable medium carries one or more programs (eg, computer-executable programs). When the one or more programs are executed by a device, the computer-readable medium implements the method of the present disclosure.

[0107] Those skilled in the art will appreciate that the above modules can be distributed in the device according to the description of the embodiment, or can be changed accordingly and only used in one or more devices different from the embodiment. The modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

[0108] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the exemplary embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several commands to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiment of the present invention.

[0109] The exemplary embodiments of the present invention are specifically shown and described above. It should be understood that the present invention is not limited to the detailed structure, configuration or implementation method described herein; on the contrary, the present invention is intended to cover various modifications and equivalent configurations included in the spirit and scope of the appended claims.

Claims

1. A privacy communication anomaly detection method, It is characterized in that include: Collecting call bill data from a communication service provider platform and preprocessing the call bill data, wherein the call bill data includes a total call volume of a private number, and the preprocessing includes constructing a feature vector according to a timestamp; Using the long short-term memory network algorithm, a detection indicator prediction model is constructed, and the dynamic threshold is calculated in real time, and the result judgment is made based on the dynamic threshold. The detection indicators include the total number of calls to the privacy number; Determine model input data according to the time period to be determined, input the determined model input data into the detection index prediction model, and obtain the index prediction value corresponding to the time period to be determined; Retrieving the actual detection index value of the time period to be judged, and judging the abnormality of the detection index according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time; According to the abnormal judgment results of the detection indicators, determine the triggering early warning mechanism.

2. The privacy communication anomaly detection method according to claim 1, It is characterized in that The abnormality judgment of the detection index is performed according to the actual detection index value, the predicted index value and the dynamic threshold calculated in real time, including: using the following expression to calculate the current moment time The ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold : ; in, Indicates the current time time The ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold value. When calculating the total call volume prediction deviation of the communication service provider platform, p is used to represent the communication service provider platform; when calculating the total call volume of the sub-platform, p is used to represent the sub-platform; The average value of the dynamic threshold of the calculated detection index in multiple specified historical time periods, the detection index includes the call volume of the communication service provider platform in the specified historical time period, the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period, and the specified historical time period includes the call volume of each day within 14 days from the current time and the current time. time The same time period, the specified historical time period is a periodic time period; For the current moment time The predicted total call volume data of the corresponding collection point; For the current moment time The actual total call volume data of the corresponding collection point; When the calculated ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold If it is not within the set threshold range, the current time is determined time The communication service provision capability of the communication service provision platform is abnormal or the communication service provision capability of the sub-platform is abnormal; When the calculated ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold When setting the threshold range, the current moment is determined time The communication service provision capability of the communication service provision platform is normal or the communication service provision capability of the sub-platform is normal.

3. The privacy communication anomaly detection method according to claim 2, It is characterized in that The step of determining abnormality of a detection index based on the actual detection index value, the predicted index value, and a dynamic threshold value calculated in real time includes: When the calculated ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold If it is not within the set threshold range, and when the number of times it is not within the set threshold range within a specific time interval is greater than the number threshold, it is determined that the communication service provision capability of the communication service provision platform or sub-platform is abnormal.

4. The privacy communication anomaly detection method according to claim 2, It is characterized in that include: The dynamic threshold of the detection index in each time period is dynamically calculated using the historical communication service business data. The dynamic threshold is used to determine whether the detection index is abnormal. Specifically, the dynamic threshold of the detection index is calculated using the following expression: = Mean(Historical_Calls, time−14days≤time′≤time−1day); in, The average value of the dynamic threshold of the calculated detection index in multiple specified historical time periods, the detection index includes the call volume of the communication service provider platform in the specified historical time period, the call volume of multiple sub-platforms that can communicate with the communication service provider platform in real time in the specified historical time period, and the specified historical time period includes the call volume of each day within 14 days from the current time. time The same time period, the specified historical time period is a periodic time period; Historical_Calls Indicates that the communication service provider platform communicates with The call volume data set with the historical timestamp in the same time period, or the call volume data of each sub-platform every day A call volume data set whose historical timestamps are in the same time period; is the total call volume sequence of the communication service provider platform or each sub-platform at a specified frequency, i.e., the collection frequency, within a specified historical period; time′ represents the total call volume sequence of the communication service provider platform every day. The historical timestamp is the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day, or the sampling time of each sub-platform every day. The historical timestamp is located at the sampling time of the call volume data set in the same time period, that is, from time-14days to time-1day.

5. The privacy communication anomaly detection method according to claim 2, It is characterized in that include: Using detection indicators of the communication service providing platform or multiple sub-platforms in multiple historical time periods in the historical communication service business data, establishing a first segmented curve trend graph and a second segmented curve trend graph corresponding to the communication service providing platform and each sub-platform respectively; the detection indicators include a total call volume; According to the obtained first segmented curve trend graph and the second segmented curve trend graph, the dynamic threshold of the detection index in each time period is calculated in real time.

6. The privacy communication anomaly detection method according to claim 1, It is characterized in that The step of determining model input data according to the time period to be determined, inputting the determined model input data into the detection index prediction model, and obtaining the index prediction value corresponding to the time period to be determined includes: According to the time period to be determined, the following data are determined to form model input data: a timestamp sequence, a historical detection indicator sequence, and a coding sequence of the communication service providing platform or sub-platform; A training data set is established to train a detection index prediction model using a timestamp sequence marked with a call volume label, a historical detection index quantity sequence, and a coding sequence of the communication service provider platform or sub-platform. The call volume label specifically refers to the total call volume corresponding to the timestamp sequence.

7. The privacy communication anomaly detection method according to claim 1, It is characterized in that Further including: Establish a detection index abnormality judgment model, A timestamp sequence marked with normal indicator labels and abnormal indicator labels, a historical detection indicator quantity sequence, and a coding sequence of the communication service provider platform or sub-platform are used to establish a training data set and train a detection indicator judgment model. The normal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is within the set threshold range, and the abnormal indicator label means that the ratio of the total call volume prediction deviation of the communication service provider platform or sub-platform to the dynamic threshold is not within the set threshold range.

8. The privacy communication anomaly detection method according to claim 1, It is characterized in that Further including: The following loss function is used to optimize the training process of the detection index prediction model: ; in, represents the difference between the total call volume prediction value output by the detection index prediction model and the actual total call volume, i.e., the loss value; Y represents the actual total call volume; f(X) represents the total call volume prediction value output by the detection index prediction model, and X is the model input data of the detection index prediction model; When the difference is less than 1%, the training of the detection index prediction model is stopped, and the trained detection index prediction model is obtained.

9. The privacy communication anomaly detection method according to claim 1, It is characterized in that Said, including: According to the abnormality judgment result of the detection indicator, specifically when the detection indicator is abnormal, it is determined that the communication service providing platform or sub-platform is abnormal, and an early warning mechanism is triggered. The early warning mechanism includes fault handling and number migration processing.

10. A privacy communication anomaly detection system, It is characterized in that It implements the privacy communication anomaly detection method according to any one of claims 1 to 9, and the privacy communication anomaly detection system comprises: A data collection module, used to collect call data from a communication service provider platform and pre-process the call data, wherein the call data includes the total call volume of a private number, and the pre-processing includes constructing a feature vector according to a timestamp; A model building module, which is used to build a detection index prediction model using a long short-term memory network algorithm, to predict the detection index, calculate the dynamic threshold in real time, and make a result judgment based on the dynamic threshold. The detection index includes the total call volume of the privacy number; A prediction module, which determines model input data according to the time period to be determined, inputs the determined model input data into the detection index prediction model, and obtains an index prediction value corresponding to the time period to be determined; A judgment module, used to retrieve the actual detection index value of the time period to be judged, and make a judgment on the abnormality of the detection index according to the actual detection index value, the index prediction value and the dynamic threshold calculated in real time; The early warning module determines the triggering of the early warning mechanism based on the abnormal judgment results of the detection indicators.