Abnormal call identification method, device, equipment and storage medium

Through big data analysis and offline modeling, abnormal phones are identified, and problems such as difficulty in detecting and updating behavioral characteristics and large investment in voice analysis equipment in the existing technology are solved, and efficient abnormal phone recognition and accuracy are achieved to adapt to changes in fraudulent behavior characteristics.

CN114567697BActive Publication Date: 2024-11-01EVERSEC BEIJING TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210192989.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-01
Publication Date
2024-11-01
Estimated Expiration
2042-03-01

AI Technical Summary

Technical Problem

The existing technology cannot update the behavioral feature detection rule base in a timely manner, and content-based detection methods require large investment in voice analysis equipment, and cannot effectively respond to the need for fraudulent means to change as criminals become more vigilant.

Method used

Using a method based on big data analysis and offline modeling, we use a single data of exceptions and normal sample sets to determine sample characteristics, filter out significant abnormal samples, train an abnormal telephone recognition model, and identify abnormal telephones in real time.

Benefits of technology

It improves the efficiency of model iteration optimization, shortens the model solidification cycle, ensures the accuracy of identification, and can effectively deal with the situation where fraudulent behavior changes with criminals' vigilance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114567697B_ABST
    Figure CN114567697B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention discloses a method, device, equipment and storage medium for identifying abnormal calls. The method includes obtaining a sample set; determining the sample characteristics of each sample, and determining significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set; screening the target abnormal sample set in the abnormal sample set according to the significant abnormal sample screening conditions, and training an abnormal call recognition model based on the target abnormal sample set; inputting the test data set obtained in real time into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set. The technical solution of the embodiment of the present invention provides a new method for abnormal call recognition based on big data analysis and offline modeling, which improves the efficiency of model iteration optimization, shortens the model solidification cycle, ensures the recognition accuracy, and can also effectively deal with the situation where the characteristics of fraudulent behavior are constantly changing as the criminals become more vigilant.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of mobile communication phone anti-fraud technology, and in particular to an abnormal phone identification method, device, equipment and storage medium. Background Art

[0002] At present, with the popularization of mobile Internet, telephone fraud cases have emerged in an endless stream of fraud methods, continuously extended fraud scenarios, and the number of frauds remains high.

[0003] Counterfeit fraud is one of the more common forms of fraud used by criminals. Through systematic scripts and clear division of labor, victims are gradually led into traps. This type of fraud is generally committed by gangs, with high professionalism and vigilance. The currently commonly used behavioral feature detection method cannot timely and targetedly update the anti-fraud rule library, while the content-based detection method is more dependent on user privacy data and requires a large investment in voice analysis equipment. The current detection method can no longer meet the anti-fraud needs that are constantly changing with the fraud methods and the criminals' increased vigilance. Summary of the invention

[0004] The embodiments of the present invention provide a method, device, equipment and storage medium for identifying abnormal phone calls, and provide a new method for realizing abnormal phone call identification based on big data analysis and offline modeling, thereby improving the efficiency of model iteration optimization, shortening the model solidification cycle, and ensuring the recognition accuracy while effectively coping with the situation in which the characteristics of fraudulent behavior change constantly as the criminals become more vigilant.

[0005] In a first aspect, an embodiment of the present invention provides a method for identifying an abnormal call, the method comprising:

[0006] Obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval;

[0007] Determine the sample characteristics of each sample based on the call record fields in each call record, and determine the significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set;

[0008] According to the significant abnormal sample screening conditions, a target abnormal sample set is screened out from the abnormal sample set, and an abnormal call recognition model is trained based on the target abnormal sample set;

[0009] The test data set acquired in real time is input into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set.

[0010] In a second aspect, an embodiment of the present invention further provides a device for identifying abnormal calls, the device comprising:

[0011] A sample set acquisition module, used to acquire a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval;

[0012] A screening condition determination module is used to determine the sample characteristics of each sample according to the call record fields in each call record, and determine the significant abnormal sample screening conditions according to the difference between the characteristics of each sample in the abnormal sample set and the normal sample set;

[0013] The model training module is used to screen the target abnormal sample set from the abnormal sample set according to the significant abnormal sample screening conditions, and train the abnormal call recognition model according to the target abnormal sample set;

[0014] The recognition result acquisition module is used to input the test data set acquired in real time into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set.

[0015] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0016] one or more processors;

[0017] A storage device for storing one or more programs;

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement a method for identifying an abnormal call as described in any embodiment of the present invention.

[0019] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, a method for identifying an abnormal telephone call as described in any embodiment of the present invention is implemented.

[0020] The embodiment of the present invention obtains a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call slip with the sample number within a preset historical time interval; according to the call slip field in each call slip, the sample feature of each sample is determined, and according to the difference between the features of each sample in the abnormal sample set and the normal sample set, a significant abnormal sample screening condition is determined; according to the significant abnormal sample screening condition, a target abnormal sample set is screened in the abnormal sample set, and an abnormal call recognition model is trained according to the target abnormal sample set; a test data set obtained in real time is input into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set, so as to solve the problem that the detection method based on behavior features in the prior art cannot timely and targetedly update the rule base, and the detection method based on call content requires a large investment in voice analysis equipment, and provides a new method for realizing abnormal call recognition based on big data analysis and offline modeling, improves the efficiency of model iteration optimization, shortens the model solidification cycle, ensures the recognition accuracy, and can also effectively deal with the situation that the fraud behavior features are constantly changing with the increasing vigilance of criminals. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of a method for identifying an abnormal call provided in Embodiment 1 of the present invention;

[0022] Figure 1a A schematic diagram of distribution characteristics of a standard deviation of the hour of a call provided by the first embodiment of the present invention;

[0023] Figure 1b A schematic diagram of distribution characteristics of the average daily total duration provided by an embodiment of the present invention;

[0024] Figure 1c A schematic diagram of a specific process of offline training of an abnormal phone call recognition model provided in the first embodiment of the present invention;

[0025] Figure 2 A schematic diagram of the structure of an abnormal telephone identification device provided in Embodiment 2 of the present invention;

[0026] Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0027] The present invention 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 invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0028] Embodiment 1

[0029] Figure 1 This is a flowchart of a method for identifying abnormal phone calls provided in Embodiment 1 of the present invention. This embodiment is applicable to the case of identifying abnormal phone numbers. The method can be performed by an abnormal phone call identification device, which can be implemented in software and / or hardware and integrated in a server. Figure 1 , the method specifically comprises the following steps:

[0030] S110, obtaining a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval.

[0031] The abnormal sample set may be abnormal telephone number data obtained from multiple third parties, for example, public case reporting data, reporting numbers obtained from operators, reporting numbers obtained from XX reporting platform, and some numbers marked by Internet labels.

[0032] The normal sample set can be obtained based on information from various cities and different number segments. A batch of seemingly normal user numbers can be randomly selected, and pre-established Internet crawler data can be used to check whether these numbers are marked as harassment or fraud (except for express delivery and real estate agencies). If there are no marks, they are regarded as normal samples, thereby constructing a normal sample set.

[0033] The preset historical time interval may refer to a time period for obtaining call records of a single sample number, and specifically may refer to obtaining call records of a single sample number in the past several consecutive days. The call record may refer to B domain call record information.

[0034] In this embodiment, after obtaining the sample set, preliminary screening can be performed by the call record fields. For example, samples with phone2_type of 5 (3-digit service number), 6 (12583 and other service numbers), 7 (securities, banking and aviation numbers) can be pre-screened out of the sample set.

[0035] S120: Determine the sample feature of each sample according to the call record fields in each call record, and determine the significant abnormal sample screening condition according to the difference between the sample features in the abnormal sample set and the normal sample set.

[0036] The sample characteristics may refer to the number of calls, the proportion of calls, the time of calls, the duration of calls, etc. The significant abnormal sample screening condition may be a condition for screening out obvious abnormal calls, and the significant abnormal sample screening condition includes at least one.

[0037] In this embodiment, the field values ​​corresponding to various fields in the sample call record can be identified, the sample characteristics of each sample can be determined, and then the sample characteristics of the abnormal samples and the sample characteristics of the normal samples can be compared to obtain the differences between the sample characteristics of the abnormal sample set and the normal sample set, thereby determining the significant abnormal sample screening conditions based on the differences.

[0038] In an optional implementation of this embodiment, determining the significant abnormal sample screening condition according to the difference between the features of each sample in the abnormal sample set and the normal sample set may include:

[0039] According to the sample features corresponding to the abnormal sample set and the normal sample set, respectively, the abnormal threshold values ​​corresponding to the sample features are calculated by the 4-quantile analysis method; according to the abnormal threshold values, the alternative abnormal sample set and the alternative normal sample set are screened from the abnormal sample set and the normal sample set; according to the differences between the sample features in the alternative abnormal sample set and the alternative normal sample set, the significant abnormal sample screening conditions are determined.

[0040] The abnormal threshold value may refer to a standard value for classifying individual samples in a sample set. The candidate abnormal sample set may be a part of abnormal samples selected from the abnormal sample set according to the abnormal threshold value of the abnormal sample set as the candidate abnormal sample set, which can be understood as filtering out normal samples that are misjudged as abnormal samples from the abnormal sample set. The candidate normal sample set may be a part of normal samples selected from the normal sample set according to the abnormal threshold value of the normal sample set as the candidate normal sample set, which can be understood as filtering out abnormal samples that are misjudged as normal samples from the normal sample set.

[0041] Specifically, the abnormal threshold values ​​corresponding to each sample feature of the abnormal sample set and the normal sample set can be calculated by the 4-quantile analysis method, so that the abnormal sample set and the normal sample set can be screened according to the abnormal threshold values ​​of each sample feature, and the misjudged samples can be screened out to obtain the alternative abnormal sample set and the alternative normal sample set. The differences between the sample features of the alternative abnormal sample set and the alternative normal sample set are compared to determine the screening conditions for significant abnormal samples. The advantage of this setting is that misjudged samples in the abnormal sample set and the normal sample set can be screened out, thereby improving the accuracy of the sample categories in the sample set.

[0042] For example, the specific implementation of the 4-quartile analysis method can be as follows:

[0043] Arrange all the values ​​corresponding to the calling duration of a batch of samples from small to large and divide them into four equal parts. The corresponding values ​​at the three dividing points are the quartiles.

[0044] The first quartile (Q1), also known as the "lower quartile", is equal to the 25th percentile of all the values ​​in the sample arranged from small to large; the second quartile (Q2), also known as the "median", is equal to the 50th percentile of all the values ​​in the sample arranged from small to large; the third quartile (Q3), also known as the "upper quartile", is equal to the 75th percentile of all the values ​​in the sample arranged from small to large. The gap between Q3 and Q1 is also called the interquartile range IQR, which includes half of all the values. IQR = Q3-Q1, reflects the degree of dispersion of the middle 50% of the data. The smaller the value, the more concentrated the data in the middle; the larger the value, the more dispersed the data in the middle.

[0045] The standard for identifying abnormal values ​​in the batch of values ​​may be: a value less than Q1-1.5IQR or greater than Q3+1.5IQR, wherein the values ​​corresponding to Q1-1.5IQR and Q3+1.5IQR are abnormal threshold values.

[0046] Optionally, determining the significant abnormal sample screening condition according to the difference between the characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set may include:

[0047] The distribution characteristics of each sample in the alternative abnormal sample set and the alternative normal sample set under each sample feature are calculated respectively, and the sample features whose difference between the distribution characteristics exceeds the preset difference threshold are selected as the target sample features; the distribution characteristics of the target sample features in the alternative abnormal sample set are used as the screening conditions for significant abnormal samples.

[0048] The distribution feature may be the distribution of the value sample feature within a period of time. The target sample feature may refer to a feature selected from multiple sample features that can be used to filter conditions. The preset difference threshold may refer to a limit value of the difference degree.

[0049] In this embodiment, the distribution of each sample in the alternative abnormal sample set and the alternative normal sample set under each sample feature can be calculated respectively, the distribution of each sample feature of the abnormal samples and the normal samples can be compared and the corresponding difference can be determined, and the sample features whose difference exceeds the preset difference threshold can be used as target sample features, so that the distribution characteristics of the target sample features in the alternative abnormal sample set can be used as significant abnormal sample screening conditions.

[0050] Take call time as an example. Figure 1a A schematic diagram of the distribution characteristics of the standard deviation of the call hour provided in Example 1 of the present invention. Within 8 days, the standard deviation of the call hour of abnormal numbers is more discrete than that of normal numbers, but the standard deviation of the call hour of most abnormal numbers within 8 days is relatively low, which indicates that the number of active hours of abnormal numbers every day is relatively similar.

[0051] Take call duration as an example. Figure 1b A schematic diagram of the distribution characteristics of the average daily total duration provided by an embodiment of the present invention. Within 8 days, the average daily total duration of abnormal numbers within 8 days is relatively long and most of them are higher than 1500 seconds, while the average daily total duration of normal numbers within 8 days is relatively short and most of them are less than 1000 seconds.

[0052] Obviously, there are obvious differences in the distribution of the standard deviation of the call hours between abnormal numbers and normal numbers, or the distribution of the average daily total duration. Therefore, the standard deviation of the call hours and / or the distribution of the average daily total duration can be used as significant abnormal sample screening conditions.

[0053] S130. According to the significant abnormal sample screening condition, a target abnormal sample set is screened in the abnormal sample set, and an abnormal call recognition model is trained based on the target abnormal sample set.

[0054] The target abnormal sample set may be an abnormal sample set used to train an abnormal call recognition model. The abnormal call recognition model may be a model used to perform abnormal call recognition.

[0055] In this embodiment, the target abnormal sample set can be screened from the abnormal sample set according to the significant abnormal sample screening condition, so as to perform model training. The advantage of such a setting is that the efficiency of model iteration optimization can be improved.

[0056] In an optional implementation of this embodiment, training an abnormal call recognition model based on a target abnormal sample set may include:

[0057] According to the target abnormal sample set, multiple candidate abnormal call recognition models constructed by machine learning algorithms are offline trained respectively; the recognition effects of the trained candidate abnormal call recognition models are compared to determine the target abnormal call recognition model; the target abnormal call recognition model is solidified to obtain the abnormal call recognition model.

[0058] Among them, the machine learning algorithm may include, for example, random forest, SVM (Support Vector Machine), GBDT (Gradient Boosting Decision Tree), lightGBM (Light Gradient Boosting Machine), and other algorithms.

[0059] For example, Figure 1cA specific flow chart of offline training of an abnormal call recognition model provided in the first embodiment of the present invention. Sample data is obtained, valid abnormal samples are screened out from all sample data according to the call slips of the sample data, and sample features of the valid abnormal samples are obtained, compared with the sample features of normal samples, and screening conditions are established to filter the valid abnormal samples to check whether the filtering results meet the expected results. If so, the model can be trained and solidified offline to obtain an abnormal call recognition model. Otherwise, the sample features can be reanalyzed.

[0060] S140, inputting the test data set acquired in real time into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set.

[0061] In an optional implementation of this embodiment, inputting the test data set acquired in real time into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set may include:

[0062] The test data set acquired in real time is input into the abnormal call recognition model to obtain an output result; the output result includes a preliminary recognition result of the abnormal calls in the test data set; according to the normal call screening conditions, the preliminary recognition result in the output result is processed to obtain the abnormal call recognition result corresponding to the test data set.

[0063] Among them, normal phone screening conditions may include screening conditions such as package value. For example, abnormal numbers generally will not be set up with high-value packages.

[0064] In this embodiment, in order to improve the accuracy of identifying abnormal calls, the preliminary identification results of the abnormal call identification model can be further screened to screen out normal numbers that are misjudged.

[0065] Optionally, the output result also includes: an early warning result for telephones with missing call list fields in the test data set;

[0066] After getting the output results, it also includes:

[0067] If the warning result exceeds the warning threshold, the test data set and the warning result are sent to the warning platform for monitoring.

[0068] In this embodiment, when the abnormal call identification model performs abnormal call identification in real time, if there is missing data in the call record field, an early warning result can be generated for the corresponding call. According to the degree of missing data, the corresponding warning degree can be included in the early warning result, and the early warning result can be sent to the early warning platform for monitoring.

[0069] Optionally, after obtaining the abnormal phone recognition result corresponding to the test data set, the method further includes:

[0070] According to the call duration feature that matches the abnormal call identification result, the abnormality degree that matches the abnormal call identification result is obtained.

[0071] Specifically, based on the final abnormal call recognition result of the test data set, the matching abnormality degree can be obtained according to the corresponding call duration feature. For example, if the caller and the called party corresponding to the abnormal call have more than 2 calls and the call duration exceeds 60 seconds, or the call duration between the called party and the caller exceeds 180 seconds, it can be judged as a deep abnormality.

[0072] The technical solution of the embodiment of the present invention is to obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call slip with the sample number within a preset historical time interval; determine the sample feature of each sample according to the call slip field in each call slip, and determine the significant abnormal sample screening condition according to the difference between the sample features of the abnormal sample set and the normal sample set; screen the target abnormal sample set in the abnormal sample set according to the significant abnormal sample screening condition, and train an abnormal call recognition model according to the target abnormal sample set; input the test data set obtained in real time into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set, thereby solving the problem in the prior art that the detection method based on behavioral features cannot timely and targetedly update the rule base, and the detection method based on call content requires a large investment in voice analysis equipment, and provides a new method for realizing abnormal call recognition based on big data analysis and offline modeling, which improves the model iteration optimization efficiency, shortens the model solidification cycle, ensures the recognition accuracy, and can also effectively deal with the situation that the fraud behavior features are constantly changing with the criminals' increased vigilance.

[0073] Embodiment 2

[0074] Figure 2 This is a schematic diagram of the structure of an abnormal phone identification device provided in the second embodiment of the present invention. The device can execute the abnormal phone identification method involved in the above embodiments. Figure 2 The device includes: a sample set acquisition module 210, a screening condition determination module 220, a model training module 230 and a recognition result acquisition module 240. Among them:

[0075] The sample set acquisition module 210 is used to acquire a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval;

[0076] A screening condition determination module 220 is used to determine the sample characteristics of each sample according to the call record fields in each call record, and determine the significant abnormal sample screening conditions according to the difference between the characteristics of each sample in the abnormal sample set and the normal sample set;

[0077] The model training module 230 is used to screen the target abnormal sample set from the abnormal sample set according to the significant abnormal sample screening condition, and train the abnormal call recognition model according to the target abnormal sample set;

[0078] The recognition result acquisition module 240 is used to input the test data set acquired in real time into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set.

[0079] The technical solution of the embodiment of the present invention is to obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call slip with the sample number within a preset historical time interval; determine the sample feature of each sample according to the call slip field in each call slip, and determine the significant abnormal sample screening condition according to the difference between the sample features of the abnormal sample set and the normal sample set; screen the target abnormal sample set in the abnormal sample set according to the significant abnormal sample screening condition, and train an abnormal call recognition model according to the target abnormal sample set; input the test data set obtained in real time into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set, thereby solving the problem in the prior art that the detection method based on behavioral features cannot timely and targetedly update the rule base, and the detection method based on call content requires a large investment in voice analysis equipment, and provides a new method for realizing abnormal call recognition based on big data analysis and offline modeling, which improves the model iteration optimization efficiency, shortens the model solidification cycle, ensures the recognition accuracy, and can also effectively deal with the situation that the fraud behavior features are constantly changing with the criminals' increased vigilance.

[0080] In the above device, optionally, the screening condition determination module 220 includes:

[0081] an abnormal threshold value calculation unit, used to calculate the abnormal threshold value corresponding to each sample feature respectively according to each sample feature respectively corresponding to the abnormal sample set and the normal sample set by using a 4-quantile analysis method;

[0082] A candidate sample set acquisition unit, configured to screen the abnormal sample set and the normal sample set to obtain a candidate abnormal sample set and a candidate normal sample set according to the abnormal threshold value;

[0083] The screening condition determination unit is used to determine the significant abnormal sample screening condition according to the difference between the characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set.

[0084] In the above device, optionally, the screening condition determination unit may be specifically used for:

[0085] Calculate the distribution characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set under each sample feature respectively, and select the sample features whose difference between the distribution characteristics exceeds the preset difference threshold as the target sample features;

[0086] The distribution characteristics of the target sample features in the candidate abnormal sample set are used as the screening conditions for significant abnormal samples.

[0087] In the above device, optionally, the model training module 230 may be used for:

[0088] According to the target abnormal sample set, multiple candidate abnormal call recognition models constructed by machine learning algorithms are trained offline respectively;

[0089] Compare the recognition effects of the trained candidate abnormal call recognition models to determine the target abnormal call recognition model;

[0090] The target abnormal call recognition model is solidified to obtain an abnormal call recognition model.

[0091] In the above device, optionally, the recognition result acquisition module 240 can be specifically used for:

[0092] Inputting the test data set acquired in real time into the abnormal phone recognition model to obtain an output result; the output result includes a preliminary recognition result of the abnormal phone calls in the test data set;

[0093] According to the normal telephone screening conditions, the preliminary recognition result in the output result is processed to obtain the abnormal telephone recognition result corresponding to the test data set.

[0094] In the above device, optionally, the output result also includes: an early warning result for a phone number with missing call list fields in the test data set;

[0095] The above device may optionally further include a monitoring module, which is used to:

[0096] If the warning result exceeds the warning threshold, the test data set and the warning result are sent to the warning platform for monitoring.

[0097] The above device may optionally further include an abnormality degree acquisition module, which is used to, after obtaining the abnormal phone recognition result corresponding to the test data set:

[0098] According to the call duration feature that matches the abnormal call identification result, the abnormality degree that matches the abnormal call identification result is obtained.

[0099] The abnormal phone call identification device provided in the embodiment of the present invention can execute the abnormal phone call identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0100] Embodiment 3

[0101] Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention is shown in FIG. Figure 3 As shown, the device includes a processor 310, a storage device 320, an input device 330 and an output device 340; the number of processors 310 in the device can be one or more. Figure 3 A processor 310 is taken as an example; the processor 310, storage device 320, input device 330 and output device 340 in the device can be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0102] The storage device 320, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the abnormal phone identification method in the embodiment of the present invention (for example, the sample set acquisition module 210, the screening condition determination module 220, the model training module 230 and the recognition result acquisition module 240 in the abnormal phone identification device). The processor 310 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the storage device 320, that is, implements the above-mentioned abnormal phone identification method, which may include:

[0103] Obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval;

[0104] Determine the sample characteristics of each sample based on the call record fields in each call record, and determine the significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set;

[0105] According to the significant abnormal sample screening conditions, a target abnormal sample set is screened out from the abnormal sample set, and an abnormal call recognition model is trained based on the target abnormal sample set;

[0106] The test data set acquired in real time is input into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set.

[0107] The storage device 320 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the storage device 320 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 storage device 320 may further include a memory remotely arranged relative to the processor 310, and these remote memories may be connected to the 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.

[0108] The input device 330 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 340 may include a display device such as a display screen.

[0109] Embodiment 4

[0110] Embodiment 4 of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to perform a method for identifying an abnormal phone call. The method may include:

[0111] Obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval;

[0112] Determine the sample characteristics of each sample based on the call record fields in each call record, and determine the significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set;

[0113] According to the significant abnormal sample screening conditions, a target abnormal sample set is screened out from the abnormal sample set, and an abnormal call recognition model is trained based on the target abnormal sample set;

[0114] The test data set acquired in real time is input into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set.

[0115] Of course, the computer-readable storage medium provided in the embodiment of the present invention stores a computer program thereon, and the computer program is not limited to the method operations described above, but can also execute the related operations in the abnormal telephone identification method provided in any embodiment of the present invention.

[0116] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means 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 such an understanding, the technical solution of the present invention, in essence, 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, a read-only memory (ROM), a random access memory (RAM), a flash memory (FLASH), a hard disk or an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0117] It is worth noting that in the embodiment of the above-mentioned abnormal telephone identification 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 the present invention.

[0118] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention 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 invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for identifying abnormal calls, characterized in that: include: Obtain a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval; Determine the sample characteristics of each sample based on the call record fields in each call record, and determine the significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set; According to the significant abnormal sample screening conditions, a target abnormal sample set is screened out from the abnormal sample set, and an abnormal call recognition model is trained based on the target abnormal sample set; Inputting the test data set acquired in real time into the abnormal call recognition model to obtain an abnormal call recognition result corresponding to the test data set; The method of determining the significant abnormal sample screening conditions based on the differences between the characteristics of each sample in the abnormal sample set and the normal sample set includes: According to each sample feature corresponding to the abnormal sample set and the normal sample set, respectively, calculating the abnormal threshold value corresponding to each sample feature by using a 4-quantile analysis method; According to the abnormal threshold value, screening out a candidate abnormal sample set and a candidate normal sample set from the abnormal sample set and the normal sample set; According to the difference between the characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set, a significant abnormal sample screening condition is determined.

2. The method according to claim 1, characterized in that According to the difference between the characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set, a significant abnormal sample screening condition is determined, including: Calculate the distribution characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set under each sample feature respectively, and select the sample features whose difference between the distribution characteristics exceeds the preset difference threshold as the target sample features; The distribution characteristics of the target sample features in the candidate abnormal sample set are used as the screening conditions for significant abnormal samples.

3. The method according to claim 1, characterized in that According to the target abnormal sample set, the abnormal call recognition model is trained, including: According to the target abnormal sample set, multiple candidate abnormal call recognition models constructed by machine learning algorithms are trained offline respectively; Compare the recognition effects of the trained candidate abnormal call recognition models to determine the target abnormal call recognition model; The target abnormal call recognition model is solidified to obtain an abnormal call recognition model.

4. The method according to claim 1, characterized in that: Inputting the test data set acquired in real time into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set includes: Inputting the test data set acquired in real time into the abnormal phone recognition model to obtain an output result; the output result includes a preliminary recognition result of the abnormal phone calls in the test data set; According to the normal telephone screening conditions, the preliminary recognition result in the output result is processed to obtain the abnormal telephone recognition result corresponding to the test data set.

5. The method according to claim 4, characterized in that The output result also includes: warning results for missing calls in the call list field in the test data set; After getting the output results, it also includes: If the warning result exceeds the warning threshold, the test data set and the warning result are sent to the warning platform for monitoring.

6. The method according to claim 4, characterized in that After obtaining the abnormal phone recognition results corresponding to the test data set, it also includes: According to the call duration feature that matches the abnormal call identification result, the abnormality degree that matches the abnormal call identification result is obtained.

7. An abnormal telephone identification device, characterized in that: include: A sample set acquisition module, used to acquire a sample set, wherein the sample set includes: an abnormal sample set and a normal sample set, each sample including a sample number and a call record with the sample number within a preset historical time interval; A screening condition determination module is used to determine the sample characteristics of each sample according to the call record fields in each call record, and determine the significant abnormal sample screening conditions according to the difference between the characteristics of each sample in the abnormal sample set and the normal sample set; The model training module is used to screen the target abnormal sample set from the abnormal sample set according to the significant abnormal sample screening conditions, and train the abnormal call recognition model according to the target abnormal sample set; The recognition result acquisition module is used to input the test data set acquired in real time into the abnormal call recognition model to obtain the abnormal call recognition result corresponding to the test data set; Wherein, the screening condition determination module includes: an abnormal threshold value calculation unit, used to calculate the abnormal threshold value corresponding to each sample feature respectively according to each sample feature respectively corresponding to the abnormal sample set and the normal sample set by using a 4-quantile analysis method; A candidate sample set acquisition unit, configured to screen the abnormal sample set and the normal sample set to obtain a candidate abnormal sample set and a candidate normal sample set according to the abnormal threshold value; The screening condition determination unit is used to determine the significant abnormal sample screening condition according to the difference between the characteristics of each sample in the candidate abnormal sample set and the candidate normal sample set.

8. An electronic device, characterized in that: The electronic device comprises: 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 method for identifying an abnormal call as described in any one of claims 1-6.

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