Card nurturing identification method and device, electronic equipment and storage medium

Abstract

This invention provides a method, apparatus, electronic device, and storage medium for identifying credit card fraud. The method includes: determining target numbers corresponding to behavioral characteristics from a set of numbers to be identified based on indicator data of a sample number set; and determining fraudulent numbers based on the target numbers. The behavioral characteristics include joint fraudulent behavior characteristics, and the target numbers corresponding to these joint fraudulent behavior characteristics are numbers corresponding to a joint fraudulent group. The joint fraudulent group is obtained by comparing the number trajectory similarity of the set of numbers to be identified with a similarity threshold corresponding to the number trajectory similarity of the set of numbers to be identified. The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and a loss function of a greedy algorithm. This invention improves the accuracy and efficiency of credit card fraud identification.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a card identification method, device, electronic device, and storage medium. Background Technology

[0002] "Card nurturing" refers to channel agents maintaining SIM cards as if they were real users in order to obtain commissions and rebates from operators. These nurturing numbers are not actually used. This practice creates a false impression of booming business, wastes limited and valuable number resources, ties up marketing resources, results in lost commissions for operators, and hinders other users from enjoying preferential policies, causing numerous problems for business operations and management. Therefore, it is necessary to identify nurturing numbers to combat this practice.

[0003] Currently, most methods rely on manually defined rules to identify card-farming activities by channel agents and determine the associated card numbers. However, these manually defined rules suffer from limitations and lag, resulting in low accuracy. Summary of the Invention

[0004] This invention provides a card maintenance identification method, device, electronic device, and storage medium to solve the problem of low accuracy in existing card maintenance identification technologies and achieve high-accuracy card maintenance identification.

[0005] This invention provides a method for identifying credit card balances, comprising:

[0006] Based on the indicator data of the sample number set, the target number corresponding to the behavioral characteristics is determined from the number set to be identified;

[0007] Based on the target number, determine the card-maintenance number;

[0008] The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified;

[0009] The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm.

[0010] According to the present invention, a method for identifying credit card fraud is provided, wherein the similarity threshold is obtained through iterative training in the following manner:

[0011] Determine the sample similarity threshold for the current iteration round;

[0012] Based on the number trajectory similarity of the sample number set and the sample similarity threshold, a first number set corresponding to the joint card-raising community and a second number set corresponding to the non-joint card-raising community are determined from the sample number set;

[0013] Based on the index distance between the first number set and the second number set, and the loss function, the sample similarity threshold is updated to obtain the sample similarity threshold for the next iteration round. The next iteration round is then used as the current iteration round until the current iteration round is the last iteration round, in order to obtain the similarity threshold. The index distance is determined based on the index data.

[0014] According to a card maintenance identification method provided by the present invention, the indicator distance includes at least one of the following: distance between the number of calls made and received, distance between data traffic, and distance between the number of days the device is powered on.

[0015] According to a method for identifying card-farming groups provided by the present invention, the joint card-farming group is determined based on the following method:

[0016] From the set of numbers to be identified, determine the first and second numbers to be compared in the current comparison round;

[0017] Based on the number of base stations of the first base station that the first number to be compared communicated with during the first preset time period, the number of base stations of the second base station that the second number to be compared communicated with during the first preset time period, and the number of base stations of the common base stations that the first number to be compared and the second number to be compared communicated with during the first preset time period, the number trajectory similarity between the first number to be compared and the second number to be compared is determined. The number trajectory similarity between the first number to be compared and the second number to be compared includes at least one of the similarity between the common base station and the first base station, the similarity between the common base station and the second base station, and the number of days the common base station appears.

[0018] The similarity of the number trajectories of the first number to be compared and the second number to be compared is compared with the similarity threshold to obtain the comparison result;

[0019] Return to the step of determining the first and second numbers to be compared in the current comparison round from the set of numbers to be identified, until the current comparison round is the last comparison round;

[0020] Based on the comparison results of the number set to be identified, the joint card-raising group is determined.

[0021] According to the card-farming identification method provided by the present invention, the target number corresponding to the channel-based behavior characteristics of the peer number is determined based on the following method:

[0022] Determine the peer number of each peer number in the set of peer numbers to be identified, and the first target channel in which the peer number of each peer number to be identified is located;

[0023] The numbers to be analyzed in the current analysis round are determined from the sets of numbers to be identified. Based on the peer numbers of each number to be identified and the first target channel, the number of peer numbers of the number to be analyzed in each channel and the percentage of peer numbers of the number to be analyzed in each channel are determined.

[0024] When both the number of numbers and the proportion of numbers meet the first preset condition, the number to be analyzed is taken as the target number corresponding to the concentrated behavior characteristics of the peer number channel. The first preset condition is determined based on the indicator data.

[0025] Return to the step of determining the number to be analyzed in the current analysis round from each set of numbers to be identified, until the current analysis round is the last analysis round.

[0026] According to the card-raising identification method provided by the present invention, the behavioral characteristics further include channel abnormal behavior characteristics, and the target number corresponding to the channel abnormal behavior characteristics is determined based on the following method:

[0027] Determine the second target channel for each number in the set of numbers to be identified;

[0028] Based on the number to be identified, the second target channel, and the second preset conditions, the target number corresponding to the abnormal behavior characteristics of the channel is determined;

[0029] The second preset condition includes at least one of the following:

[0030] If the number of first numbers with the same International Mobile Equipment Identity (IMEI) on the same channel exceeds a first preset threshold, the first number is the target number corresponding to the channel's abnormal behavior characteristics.

[0031] If the number of second numbers on the same channel and at the same base station exceeds the second preset threshold, the second number is the target number corresponding to the abnormal behavior characteristics of the channel;

[0032] If the number of third numbers on the same channel and with the same first peer number is greater than a third preset threshold, the third number is the target number corresponding to the abnormal behavior characteristics of the channel, and the first peer number is the number that the third number communicates with the most.

[0033] If the number of fourth numbers that are on the same channel and have mutual calling behavior is greater than the fourth preset threshold and greater than the fifth preset threshold, then the fourth number is the target number corresponding to the abnormal behavior characteristics of the channel. The fifth preset threshold is obtained by multiplying the target number release volume by a preset percentage. The target number release volume is the number of numbers released by the channel where the fourth number is located in the second preset time period.

[0034] The first preset threshold, the second preset threshold, the third preset threshold, the fourth preset threshold, and the preset percentage are determined based on the indicator data.

[0035] According to the card-maintenance identification method provided by the present invention, the behavioral characteristics further include silent behavior characteristics, and the target number corresponding to the silent behavior characteristics is determined based on the following method:

[0036] Determine the outgoing call duration, incoming call duration, and data usage for each number in the set of numbers to be identified;

[0037] From the set of numbers to be identified, determine the fifth number whose call duration is less than or equal to a preset call duration threshold, whose call duration is less than or equal to a preset call duration threshold, and whose data usage is less than or equal to a preset data usage threshold, and use the fifth number as the target number corresponding to the silent behavior feature;

[0038] The preset outgoing call duration threshold, the preset incoming call duration threshold, and the preset traffic threshold are determined based on the indicator data.

[0039] According to the card-farming identification method provided by the present invention, the behavioral characteristics further include normal number behavioral characteristics, and the target number corresponding to the normal number behavioral characteristics is determined based on the following method:

[0040] Determine the average monthly recharge amount for each number in the set of numbers to be identified;

[0041] From the set of numbers to be identified, the sixth number whose average monthly recharge amount is greater than a preset amount threshold is determined, and the sixth number is used as the target number corresponding to the normal number behavior characteristics.

[0042] The preset amount threshold is determined based on the indicator data.

[0043] According to the present invention, a method for identifying card farming is provided, wherein the target number includes abnormal numbers and numbers corresponding to normal number behavior characteristics, and the abnormal numbers include numbers corresponding to the joint card farming group, numbers corresponding to the concentrated behavior characteristics of the counterpart number channel, numbers corresponding to the abnormal behavior characteristics of the channel, and numbers corresponding to the silent behavior characteristics;

[0044] The process of determining the card-maintenance number based on the target number includes:

[0045] Determine the union of the numbers corresponding to the joint card-raising group, the numbers corresponding to the channel-based concentrated behavior characteristics of the counterpart numbers, the numbers corresponding to the channel-based abnormal behavior characteristics, and the numbers corresponding to the silent behavior characteristics;

[0046] Based on the numbers corresponding to the normal number behavior characteristics, the normal numbers in the union set are removed to obtain the card-maintenance numbers.

[0047] According to the present invention, a method for identifying credit card fraud is provided, wherein the indicator data includes at least one of the following: user basic information, call behavior, data services, base station information, consumption characteristics, social circle, channel characteristics, and power on / off characteristics.

[0048] The present invention also provides a card recognition device, comprising:

[0049] The first determination module is used to determine the target number corresponding to the behavioral characteristics from the set of numbers to be identified based on the indicator data of the sample number set;

[0050] The second determining module is used to determine the card maintenance number based on the target number;

[0051] The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified;

[0052] The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm.

[0053] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the card identification method described above.

[0054] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the card identification method as described above.

[0055] The present invention provides a method, apparatus, electronic device, and storage medium for identifying card-farming behavior. Based on indicator data from a sample number set, it determines the target number corresponding to behavioral characteristics from the set of numbers to be identified; based on the target number, it determines the card-farming number. Through this method, card-farming identification and determination of the card-farming number can be performed. The behavioral characteristics include joint card-farming behavior characteristics. The joint card-farming group is obtained by comparing the similarity of the number trajectory of the set of numbers to be identified with a similarity threshold corresponding to the similarity of the number trajectory of the set of numbers to be identified, thereby determining the number corresponding to the joint card-farming group as the target number. Furthermore, the similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, indicator data, and a loss function of a greedy algorithm. This allows for automatic training based on number trajectory similarity to obtain the similarity threshold, eliminating the need for manually defining card-farming identification rules and improving the accuracy and efficiency of card-farming identification. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0057] Figure 1 This is one of the flowcharts illustrating the card recognition method provided by the present invention;

[0058] Figure 2 The second flowchart illustrates the card recognition method provided by this invention.

[0059] Figure 3 This is a schematic diagram of the structure of the card recognition device provided by the present invention;

[0060] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0062] Figure 1 This is one of the flowcharts illustrating the card recognition method provided by the present invention, such as... Figure 1 As shown, the card identification method includes:

[0063] Step 110: Based on the indicator data of the sample number set, determine the target number corresponding to the behavioral characteristics from the number set to be identified.

[0064] Here, the sample number set includes multiple sample numbers, which are obtained from historical number data. The indicator data of this sample number set includes the indicator data of each sample number in the sample number set.

[0065] The data in this indicator may include, but is not limited to, at least one of the following: basic user information, call behavior, data services, base station information, consumption characteristics, social circles, channel characteristics, and power on / off characteristics, etc.

[0066] The user's basic information may include, but is not limited to: network access duration, activation channel, activation time, etc. The call behavior may include, but is not limited to: call duration (both outgoing and incoming), number of calls (both outgoing and incoming), roaming duration, etc. The data services may include, but are not limited to: data usage, frequently used apps and data usage, etc. The base station information may include, but is not limited to: longest-stayed base station, number of base stations, etc. The consumption characteristics may include, but are not limited to: consumption amount (average monthly consumption amount, average annual consumption amount, etc.), recharge amount (average monthly recharge amount, average annual recharge amount, etc.). The social circle may include, but is not limited to: number of people in the social circle, the other party's phone number, the channel the other party's phone number is on, etc. The power on / off characteristics may include, but are not limited to: number of days the phone is on, duration the phone is on, etc.

[0067] The indicator data of this sample number set is used for behavioral feature analysis to obtain the judgment conditions and judgment thresholds for each target number. Specifically, based on the indicator data of the sample number set, behavioral feature analysis is performed to obtain the preset conditions or preset thresholds for the target numbers corresponding to the behavioral features; then, based on the preset conditions or preset thresholds, the target numbers corresponding to the behavioral features are determined from the set of numbers to be identified.

[0068] It should be noted that the indicator data on which the preset conditions or preset thresholds for determining the target numbers corresponding to each behavioral characteristic are based can be set according to actual needs. Specifically, various indicators are combined to find the combination feature rules of abnormal numbers with the highest weight in the circle. Each behavioral characteristic can correspond to a combination feature rule. For example, the combination features corresponding to the joint card-farming behavior characteristics include base station information, the number of outgoing and incoming calls, data traffic, and the number of days the device has been powered on.

[0069] Here, behavioral characteristics may include, but are not limited to, at least one of the following: joint card-farming behavior characteristics, abnormal channel behavior characteristics, silent behavior characteristics, concentrated behavior characteristics of the counterpart number channel, normal number behavior characteristics, etc. Correspondingly, target numbers may include, but are not limited to, at least one of the following: target numbers corresponding to joint card-farming behavior characteristics, target numbers corresponding to abnormal channel behavior characteristics, target numbers corresponding to silent behavior characteristics, target numbers corresponding to concentrated behavior characteristics of the counterpart number channel, target numbers corresponding to normal number behavior characteristics, etc.

[0070] Furthermore, the target numbers corresponding to the behavioral characteristics can be divided into abnormal numbers and normal numbers. Abnormal numbers are numbers suspected of being used for credit card farming, while normal numbers are numbers not used for credit card farming. In other words, normal numbers can exclude suspected credit card farming numbers.

[0071] Step 120: Determine the card maintenance number based on the target number.

[0072] Here, the target number may include one or more, and correspondingly, the card-maintenance number may also include one or more. Both the target number and the card-maintenance number are numbers in the set of numbers to be identified.

[0073] In some embodiments, if the target number only includes abnormal numbers, the union of the abnormal numbers is determined, and each number in the union is identified as the card-maintenance number. If the abnormal numbers only include numbers corresponding to one behavioral characteristic, then the number corresponding to that behavioral characteristic is directly used as the card-maintenance number; if the abnormal numbers include numbers corresponding to multiple behavioral characteristics, then the union of the numbers corresponding to each behavioral characteristic is determined, and each number in the union is used as the card-maintenance number.

[0074] In other embodiments, when the target number includes both abnormal and normal numbers, the union of the abnormal numbers is determined, and the normal numbers in the union are removed to obtain the card maintenance number. If the abnormal numbers include only one type of behavioral characteristic, then the numbers corresponding to that behavioral characteristic are directly used as the union; if the abnormal numbers include numbers corresponding to multiple behavioral characteristics, then the union of the numbers corresponding to each behavioral characteristic is determined.

[0075] In one embodiment, the target number includes abnormal numbers and numbers corresponding to normal number behavior characteristics. The abnormal numbers include at least one of the following: numbers corresponding to a joint card-farming group, numbers corresponding to the concentrated behavior characteristics of the counterpart number channel, numbers corresponding to the abnormal behavior characteristics of the channel, and numbers corresponding to the silent behavior characteristics. Specifically, the union of each abnormal number is determined, and the normal numbers in the union are removed based on the numbers corresponding to the normal number behavior characteristics to obtain the card-farming number.

[0076] The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified;

[0077] The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm.

[0078] Here, a "joint card-farming group" refers to a set of phone numbers that are used together to farm credit cards. For example, if 10 phone numbers are used for joint card farming, then the set of these 10 phone numbers is called a "joint card-farming group."

[0079] Here, the similarity of the number trajectories of the set of numbers to be identified is determined based on the base station information of the base stations communicated by each number in the set. Correspondingly, the similarity of the number trajectories of the sample number set is also determined based on the base station information of the base stations communicated by each sample number in the sample number set.

[0080] Specifically, if the similarity of the number trajectory of two numbers in the set of numbers to be identified is greater than the similarity threshold corresponding to the number trajectory similarity of the set of numbers to be identified, then the two numbers to be identified are identified as the same joint card-raising group.

[0081] The similarity of the phone number trajectories in this sample number set is used to determine whether the sample number set is a joint card-farming community or not. Specifically, the similarity of the phone number trajectories in the sample number set is compared with a sample similarity threshold, and the joint card-farming community and not are determined based on the comparison results.

[0082] Here, the indicator data of the sample number set is used to determine the distance between joint card-raising communities and non-joint card-raising communities, and then the sample similarity threshold is iteratively updated based on the distance and the loss function of the greedy algorithm.

[0083] In one embodiment, the greedy algorithm can be the Fastgreedy algorithm. The Fastgreedy algorithm is a method for finding the optimal solution to a problem. This method typically divides the solution process into several steps, each employing a greedy principle to select the best / optimal choice (the locally most advantageous choice) under the current state, hoping that the final result is also the best / optimal solution. Specifically, based on the Fastgreedy algorithm and considering the application scenario of user communication, the base station information where the user resides is viewed as a community. In each iteration, the Q-value of each user's base station is calculated, and the two communities with the largest Q-values ​​are merged until the entire network is merged into one community. More specifically, the entire process can be represented as a tree diagram, from which the level with the largest Q-value is selected to obtain the final community structure. It is understood that using the Fastgreedy algorithm can eliminate exhaustive operations, thus more simply and efficiently determining the similarity threshold. This allows the network community graph to be constructed using the trajectory data of a number roaming across multiple base stations, enhancing the accuracy of identifying card-farming behavior.

[0084] The card-farming identification method provided in this invention determines the target number corresponding to the behavioral characteristics from the set of numbers to be identified based on the indicator data of the sample number set; and determines the card-farming number based on the target number. Through this method, card-farming identification and identification of card-farming numbers can be performed. The behavioral characteristics include joint card-farming behavioral characteristics. The joint card-farming group is obtained by comparing the similarity of the number trajectory of the set of numbers to be identified with the similarity threshold corresponding to the similarity of the number trajectory of the set of numbers to be identified, thereby determining the number corresponding to the joint card-farming group as the target number. Furthermore, the similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, indicator data, and the loss function of a greedy algorithm. This allows for automatic training based on the number trajectory similarity to obtain the similarity threshold, eliminating the need for manually defining card-farming identification rules and improving the accuracy and efficiency of card-farming identification.

[0085] Based on the above embodiments, Figure 2 This is a second flowchart illustrating the card recognition method provided by the present invention, as shown below. Figure 2 As shown, the similarity threshold is obtained through iterative training in the following manner:

[0086] Step 210: Determine the sample similarity threshold for the current iteration round.

[0087] It should be noted that the iterative training process includes multiple iterations, and the specific number of iterations can be set according to actual needs. In one embodiment, the number of iterations can be directly set to a preset number, such as 1200; in another embodiment, the iteration terminates when the loss function converges to its minimum value.

[0088] If the current iteration is the first, the corresponding sample similarity threshold is the initial threshold. For example, if the initial sample similarity threshold is 0, it can be understood that the value of the loss function obtained at this time is infinite.

[0089] If the current iteration is not the first iteration, the corresponding sample similarity threshold is the similarity threshold obtained from the previous iteration. In one specific embodiment, the learning rate is set to 0.1, and the sample similarity threshold for the current iteration is obtained by adding 0.1 to the sample similarity threshold of the previous iteration.

[0090] It should be noted that the number of sample similarity thresholds is the same as the number of number trajectory similarities in the sample number set.

[0091] Since the number trajectory similarity of the sample number set includes the number trajectory similarity of any two sample numbers, this explanation uses two sample numbers (the first sample number and the second sample number) as an example. In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the similarity between a common base station and the first base station, then the sample similarity threshold includes the sample similarity threshold corresponding to the similarity between the common base station and the first base station. In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the similarity between a common base station and the second base station, then the sample similarity threshold includes the sample similarity threshold corresponding to the similarity between the common base station and the second base station. In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the number of days the common base station appears, then the sample similarity threshold includes the sample similarity threshold corresponding to the number of days the common base station appears. The first base station is the base station that the first sample number communicates with during the first preset time period; the second base station is the base station that the second sample number communicates with during the first preset time period; the common base station is the same base station that the first sample number and the second sample number communicate with during the first preset time period; the first preset time period can be set according to actual needs, such as two weeks, thereby shortening the identification cycle to one week and improving the timeliness of card maintenance identification.

[0092] Step 220: Based on the number trajectory similarity of the sample number set and the sample similarity threshold, determine the first number set corresponding to the joint card-raising community and the second number set corresponding to the non-joint card-raising community from the sample number set.

[0093] It should be noted that using a greedy algorithm to select number pairs with high trajectory similarity from the sample number set requires dividing the sample number set into joint card-raising communities and non-joint card-raising communities. In one specific embodiment, numbers with trajectory similarity greater than or equal to the sample similarity threshold are added to the joint card-raising community, while numbers with trajectory similarity less than the sample similarity threshold are added to the non-joint card-raising community.

[0094] This explanation uses two sample numbers (the first sample number and the second sample number) as an example.

[0095] In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the similarity between the common base station and the first base station, and the sample similarity threshold includes the sample similarity threshold corresponding to the similarity between the common base station and the first base station, then if the similarity between the common base station and the first base station is greater than or equal to the sample similarity threshold corresponding to the similarity between the common base station and the first base station, the first sample number and the second sample number are added to the joint card maintenance community.

[0096] In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the similarity between the common base station and the second base station, and the sample similarity threshold includes the sample similarity threshold corresponding to the similarity between the common base station and the second base station, then if the similarity between the common base station and the second base station is greater than or equal to the sample similarity threshold corresponding to the similarity between the common base station and the second base station, the first sample number and the second sample number are added to the joint card maintenance community.

[0097] In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the number of days the common base station appears, and the sample similarity threshold includes the sample similarity threshold corresponding to the number of days the common base station appears, then if the number of days the common base station appears is greater than or equal to the sample similarity threshold corresponding to the number of days the common base station appears, the first sample number and the second sample number are added to the joint card maintenance community.

[0098] In one embodiment, if the number trajectory similarity of the first sample number and the second sample number includes the similarity between the common base station and the first base station, the similarity between the common base station and the second base station, and the number of days the common base station appears, and the sample similarity threshold includes the sample similarity threshold corresponding to the similarity between the common base station and the first base station, the sample similarity threshold corresponding to the similarity between the common base station and the second base station, and the sample similarity threshold corresponding to the number of days the common base station appears, then if the similarity between the common base station and the first base station is greater than or equal to the sample similarity threshold corresponding to the similarity between the common base station and the first base station, and the similarity between the common base station and the second base station is greater than or equal to the sample similarity threshold corresponding to the similarity between the common base station and the second base station, and the number of days the common base station appears is greater than or equal to the sample similarity threshold corresponding to the number of days the common base station appears, then the first sample number and the second sample number are added to the joint card maintenance community.

[0099] The first base station is the base station that the first sample number communicates with during the first preset time period; the second base station is the base station that the second sample number communicates with during the first preset time period; the common base station is the same base station that the first sample number and the second sample number communicate with during the first preset time period; the first preset time period can be set according to actual needs, such as two weeks, thereby shortening the identification cycle to one week and improving the timeliness of card maintenance identification.

[0100] Step 230: Based on the index distance between the first number set and the second number set, and the loss function, update the sample similarity threshold to obtain the sample similarity threshold for the next iteration round, and use the next iteration round as the current iteration round until the current iteration round is the last iteration round, so as to obtain the similarity threshold. The index distance is determined based on the index data.

[0101] It should be noted that, according to the evaluation principles of the community algorithm, the distance between sample numbers within the same community should be as small as possible, while the distance between sample numbers between communities should be as large as possible. Based on this, the index distance between the first number set and the second number set is determined.

[0102] Here, the indicator distance is used to characterize the distance between the first number set and the second number set, that is, to characterize the distance between the joint card-raising community and the non-joint card-raising community. This indicator distance needs to be determined based on indicators, which can be determined based on indicator data from the sample number set. The indicator distance can include, but is not limited to, at least one of the following: the distance between the number of calls made and received, the distance in terms of data traffic, and the distance between the number of days the device is active, etc. In one embodiment, the indicator distance can be characterized using Euclidean distance.

[0103] Furthermore, it should be noted that the greater the distance between the indicators of two communities, and the smaller the reciprocal of their indicator distance, the greater the distinguishability between the two communities, and the more accurate the division of the two communities. Based on this, the reciprocal of the indicator distance between joint card-holding communities and non-joint card-holding communities can be used as a basis for judging whether the community division algorithm is reasonable, that is, judging whether the joint card-holding communities and non-joint card-holding communities divided in step 220 above are accurate. Based on this, the reciprocal of the indicator distance is used as the loss function of the greedy algorithm.

[0104] In one embodiment, the metric distance includes the distance of the number of calls made and called, and the loss function is the reciprocal of the distance of the number of calls made. In another embodiment, the metric distance includes the distance of data traffic, and the loss function is the reciprocal of the distance of data traffic. In yet another embodiment, the metric distance includes the distance of the number of days the device is powered on, and the loss function is the reciprocal of the distance of the number of days the device is powered on.

[0105] In another embodiment, the indicator distance includes the distance between the number of calls, the distance of traffic, and the distance of the number of days the device is powered on. First, the reciprocal of the distance between the number of calls, the reciprocal of the distance of traffic, and the reciprocal of the distance of the number of days the device is powered on are determined. Then, the three reciprocals are fused to obtain the loss function.

[0106] For ease of understanding, please refer to the following formula:

[0107]

[0108] In the formula, Loss represents the loss function; avg() represents the averaging operation; A1 represents the number of calls initiated and received by each sample number in the joint card-raising community, that is, the number of calls initiated and received by each sample number in the first number set, and avg(A1) represents the average number of calls initiated and received by each sample number in the joint card-raising community. For example, if the joint card-raising community includes 10 sample numbers, and the number of calls initiated and received by these 10 sample numbers are 8, 10, 13, 7, 12, 13, 7, 12, 8, and 10 respectively, then avg(A1) is 10; B1 represents the number of calls initiated and received by each sample number in the non-joint card-raising community, that is, the number of calls initiated and received by each sample number in the second number set, and avg(B1) represents the average number of calls initiated and received by each sample number in the non-joint card-raising community. For example, if the non-joint card-raising community includes 10 sample numbers, and the number of calls initiated and received by these 10 sample numbers are 9, 11, 14, 8, 13, 14, 8, and 10 respectively, then avg(A1) is 10. Given 13, 9, and 11, avg(B1) is 11; A2 represents the usage traffic of each sample number in the joint card-raising community, i.e., the usage traffic of each sample number in the first number set, and avg(A2) represents the average usage traffic of each sample number in the joint card-raising community; B2 represents the usage traffic of each sample number in the non-joint card-raising community, i.e., the usage traffic of each sample number in the second number set, and avg(B2) represents the average usage traffic of each sample number in the non-joint card-raising community; A3 represents the number of days the phone is powered on in the joint card-raising community, i.e., the number of days the phone is powered on in the first number set, and avg(A3) represents the average number of days the phone is powered on in the joint card-raising community; B3 represents the number of days the phone is powered on in the non-joint card-raising community, i.e., the number of days the phone is powered on in the second number set, and avg(B3) represents the average number of days the phone is powered on in the non-joint card-raising community.

[0109] It should be noted that the division between joint card-raising communities and non-joint card-raising communities, i.e. the determination of joint card-raising groups, is based on a similarity threshold. Therefore, to ensure the accuracy of the determination of joint card-raising groups, the sample similarity threshold needs to be updated, and the optimal similarity threshold is finally obtained through iterative training.

[0110] It is understandable that updating the sample similarity threshold yields the sample similarity threshold for the next iteration. At this point, the joint card-raising community and the non-joint card-raising community determined by the sample similarity threshold of the next iteration are different from the joint card-raising community and the non-joint card-raising community of the current iteration. Accordingly, the loss function values ​​obtained are also different.

[0111] It is understandable that the loss function value corresponding to the sample similarity threshold of the current iteration is continuously solved iteratively, and then the sample similarity threshold corresponding to the minimum value of the loss function is determined, so that the sample similarity threshold corresponding to the minimum value is determined as the final similarity threshold.

[0112] The card-farming identification method provided in this invention divides the community into joint card-farming communities and non-joint card-farming communities by using a sample similarity threshold. Then, based on the index distance between the joint card-farming communities and non-joint card-farming communities, the loss function of the greedy algorithm is solved. Based on the value of the loss function, the sample similarity threshold is updated, thereby continuously iterating and training the sample similarity threshold to obtain the optimal similarity threshold. This improves the accuracy of identifying joint card-farming groups and ultimately enhances the accuracy of card-farming identification.

[0113] Based on any of the above embodiments, in this method, the indicator distance includes at least one of the distance between the number of calls made and the number of data traffic, and the distance between the number of days the device is powered on.

[0114] Here, the number of calls is the sum of the number of calls made and the number of calls received. In one embodiment, the distance between the number of calls made and the number of calls received in the first number set and the second number set is the distance between the average number of calls made and the average number of calls made and the average number of calls made and the average number of calls made and the average number of calls made in the second number set.

[0115] Here, traffic refers to usage traffic. In one embodiment, the distance between the traffic of the first number set and the second number set is the distance between the average usage traffic of each sample number in the first number set and the average usage traffic of each sample number in the second number set.

[0116] Here, the number of days the device is powered on refers to the number of days the terminal containing the number has been powered on. In one embodiment, the distance between the number of days the first number set and the second number set is the distance between the average number of days the sample numbers in the first number set are powered on and the average number of days the sample numbers in the second number set are powered on.

[0117] The card-farming identification method provided in this invention can further improve the training effect of similarity threshold by using real-time indicator data such as the number of calls made by the caller and the recipient, traffic, and the number of days the device is powered on, thereby improving the accuracy of identifying joint card-farming groups and ultimately improving the accuracy of card-farming identification.

[0118] Based on any of the above embodiments, in this method, the joint card-raising group is determined in the following way:

[0119] From the set of numbers to be identified, determine the first and second numbers to be compared in the current comparison round;

[0120] Based on the number of base stations of the first base station that the first number to be compared communicated with during the first preset time period, the number of base stations of the second base station that the second number to be compared communicated with during the first preset time period, and the number of base stations of the common base stations that the first number to be compared and the second number to be compared communicated with during the first preset time period, the number trajectory similarity between the first number to be compared and the second number to be compared is determined. The number trajectory similarity between the first number to be compared and the second number to be compared includes at least one of the similarity between the common base station and the first base station, the similarity between the common base station and the second base station, and the number of days the common base station appears.

[0121] The similarity of the number trajectories of the first number to be compared and the second number to be compared is compared with the similarity threshold to obtain the comparison result;

[0122] Return to the step of determining the first and second numbers to be compared in the current comparison round from the set of numbers to be identified, until the current comparison round is the last comparison round;

[0123] Based on the comparison results of the number set to be identified, the joint card-raising group is determined.

[0124] It should be noted that it is necessary to filter out number pairs with high trajectory similarity from the set of numbers to be identified. Therefore, in each comparison round, the first and second numbers to be compared must be determined. Furthermore, when the set of numbers to be identified includes more than two numbers, multiple comparisons are required; therefore, it is necessary to determine the first and second numbers to be compared in the current comparison round.

[0125] Here, the first preset time period can be set according to actual needs, such as 1 week or 2 weeks. The first base station communicated with indicates that the first number to be compared has interacted with this first base station. The number of base stations communicated with by the first number to be compared within the first preset time period may include, but is not limited to: the number of base stations communicated with by the first number to be compared each day within the first preset time period, the total number of base stations communicated by the first number to be compared within the first preset time period, etc. The second base station communicated with indicates that the second number to be compared has interacted with this second base station. The number of base stations communicated with by the second number to be compared within the first preset time period may include, but is not limited to: the number of base stations communicated with by the second number to be compared each day within the first preset time period, the total number of base stations communicated by the second number to be compared within the first preset time period, etc.

[0126] The common base station for communication is a base station that both the first number to be compared and the second number to be compared have communicated with. The number of common base stations that the first number to be compared and the second number to be compared communicate with within the first preset time period may include, but is not limited to: the number of the same base stations that the first number to be compared and the second number to be compared communicate with each day within the first preset time period, and the number of all the same base stations that the first number to be compared and the second number to be compared communicate with within the first preset time period.

[0127] Here, the similarity between the shared base station and the first base station represents the degree of similarity between the shared base station of the first and second numbers to be compared and all base stations of the first number to be compared. The similarity between the shared base station and the second base station represents the degree of similarity between the shared base station of the first and second numbers to be compared and all base stations of the second number to be compared. The number of days the shared base station appeared represents the number of days the first and second numbers to be compared appeared at the shared base station.

[0128] In one specific embodiment, based on the number of base stations to which the first number to be compared communicates within a first preset time period, the number of days during which the first number to be compared has base station records within the first preset time period is determined. Then, based on the number of days during which the first number to be compared has base station records, the number of base stations to which the first number to be compared communicates each day within the first preset time period, and the number of the same base stations to which the first and second numbers to be compared communicate each day within the first preset time period, the similarity between the common base station and the first base station is determined. Based on the number of base stations to which the second number to be compared communicates within the first preset time period, the number of days during which the second number to be compared has base station records within the first preset time period is determined. Then, based on the number of days during which the second number to be compared has base station records, the number of base stations to which the second number to be compared communicates each day within the first preset time period, and the number of the same base stations to which the first and second numbers to be compared communicate each day within the first preset time period, the similarity between the common base station and the second base station is determined. Based on the number of common base stations to which the first and second numbers to be compared communicate within the first preset time period, the number of days the common base station appears is determined.

[0129] For ease of understanding, a specific embodiment is described here, referring to the following formula:

[0130]

[0131]

[0132] In the formula, Similarity_a represents the similarity between the shared base station and the first base station; a represents the first number to be compared, and b represents the second number to be compared; sim_day i (a,b) represents the number of base stations that the first number to be compared and the second number to be compared communicate with on the i-th day within the first preset time period. This represents the number of base stations that the first number to be compared communicated with on day i within the first preset time period; day a This indicates the number of days the first number to be compared has been recorded by a base station within the first preset time period; Similarity_b indicates the similarity between the shared base station and the second base station; This indicates the number of base stations that the second number to be compared communicated with on day i within the first preset time period; day b This indicates the number of days the second number to be compared was recorded by a base station within the first preset time period.

[0133] For example, the first preset time period is from April 17th to April 22nd. For ease of understanding, please refer to the table below:

[0134]

[0135]

[0136] Based on the table above, the following results are obtained:

[0137]

[0138]

[0139] sameday(a,b) = 5;

[0140] Here, sameday(a,b) represents the number of days that the shared base station appears.

[0141] Specifically, if the similarity of the number trajectory of the first number to be compared and the second number to be compared is greater than or equal to the similarity threshold (i.e., the comparison result of the current comparison round), the first number to be compared and the second number to be compared are determined as a number pair with high trajectory similarity.

[0142] If the number trajectory similarity includes multiple parameters, all parameters must be greater than or equal to the corresponding similarity threshold in order to determine the first and second numbers to be compared as a number pair with high trajectory similarity.

[0143] In one embodiment, if the similarity between the common base station and the first base station is greater than or equal to the similarity threshold corresponding to the similarity between the common base station and the first base station, and the similarity between the common base station and the second base station is greater than or equal to the similarity threshold corresponding to the similarity between the common base station and the second base station, and the number of days the common base station appears is greater than or equal to the similarity threshold corresponding to the number of days the common base station appears, then the first number to be compared and the second number to be compared are determined as a number pair with high trajectory similarity.

[0144] Here, a joint card-farming group can include one or more groups. Any two number pairs within the same joint card-farming group are number pairs with high trajectory similarity.

[0145] In one embodiment, smaller joint card-farming groups are removed from all joint card-farming groups. For example, joint card-farming groups with fewer than 10 numbers are considered small joint card-farming groups.

[0146] Specifically, a social network topology map of the set of numbers to be identified is constructed, and based on the comparison results of the set of numbers to be identified, the joint card-farming group is determined from the social network topology map.

[0147] The card-farming identification method provided in this invention determines the similarity of the number trajectories of each number pair in the number set by using the movement trajectory data of the number set to be identified between base stations. Then, based on the comparison result of the number trajectory similarity and the similarity threshold, it identifies the joint card-farming group, providing support for the identification of the joint card-farming group and ultimately improving the identification accuracy of card-farming identification.

[0148] Based on any of the above embodiments, the behavioral characteristics further include remote number channel concentration behavioral characteristics, and the target number corresponding to the remote number channel concentration behavioral characteristics is determined based on the following method:

[0149] Determine the peer number of each peer number in the set of peer numbers to be identified, and the first target channel in which the peer number of each peer number to be identified is located;

[0150] The numbers to be analyzed in the current analysis round are determined from the sets of numbers to be identified. Based on the peer numbers of each number to be identified and the first target channel, the number of peer numbers of the number to be analyzed in each channel and the percentage of peer numbers of the number to be analyzed in each channel are determined.

[0151] When both the number of numbers and the proportion of numbers meet the first preset condition, the number to be analyzed is taken as the target number corresponding to the concentrated behavior characteristics of the peer number channel. The first preset condition is determined based on the indicator data.

[0152] Return to the step of determining the number to be analyzed in the current analysis round from each set of numbers to be identified, until the current analysis round is the last analysis round.

[0153] Here, each number to be identified can correspond to one or more peer numbers, and each peer number corresponds to a target channel.

[0154] Here, the number of the peer numbers of the number to be analyzed across each channel includes the total number of peer numbers on each channel. The percentage of the peer numbers of the number to be analyzed across each channel includes the percentage of peer numbers on each channel out of all peer numbers.

[0155] Here, the first preset condition is obtained by clustering analysis of the indicator data of the sample number set. In one embodiment, clustering analysis is performed on features such as call behavior, data services, and power on / off status, and the first preset condition is determined based on the clustering analysis results.

[0156] The first preset condition includes a number count threshold and a number percentage threshold. Specifically, the first preset condition includes determining the number to be analyzed as the target number corresponding to the channel concentration behavior characteristic of the target number if the number of its counterpart numbers concentrated on a certain channel is greater than the number count threshold, and the percentage of counterpart numbers concentrated on that certain channel is greater than the number percentage threshold. For example, if the number count threshold is 6 and the number percentage threshold is 60%, then the target number corresponding to the channel concentration behavior characteristic has more than 6 counterpart numbers concentrated on a certain channel, and the counterpart numbers on that certain channel account for more than 60% of all counterpart numbers of the target number.

[0157] Based on the above, the number of phone numbers threshold and the number percentage threshold are obtained by clustering analysis of the indicator data of the sample number set. In one embodiment, clustering analysis is performed on features such as call behavior, data services, and power on / off status, and the number of phone numbers threshold and the number percentage threshold are determined based on the clustering analysis results.

[0158] The card-farming identification method provided in this invention determines the target number corresponding to the channel-specific behavior characteristics of the peer number by analyzing the number of peer numbers in each channel and the proportion of peer numbers in each channel. This can improve the accuracy of card-farming identification by adding channel-specific behavior characteristics of peer numbers to scenarios involving joint channel card farming.

[0159] Based on any of the above embodiments, in this method, the behavioral characteristics further include channel abnormal behavior characteristics, and the target number corresponding to the channel abnormal behavior characteristics is determined based on the following method:

[0160] Determine the second target channel for each number in the set of numbers to be identified;

[0161] Based on the number to be identified, the second target channel, and the second preset conditions, the target number corresponding to the abnormal behavior characteristics of the channel is determined;

[0162] The second preset condition includes at least one of the following:

[0163] If the number of first numbers with the same International Mobile Equipment Identity (IMEI) on the same channel exceeds a first preset threshold, the first number is the target number corresponding to the channel's abnormal behavior characteristics.

[0164] If the number of second numbers on the same channel and at the same base station exceeds the second preset threshold, the second number is the target number corresponding to the abnormal behavior characteristics of the channel;

[0165] If the number of third numbers on the same channel and with the same first peer number is greater than a third preset threshold, the third number is the target number corresponding to the abnormal behavior characteristics of the channel, and the first peer number is the number that the third number communicates with the most.

[0166] If the number of fourth numbers that are on the same channel and have mutual calling behavior is greater than the fourth preset threshold and greater than the fifth preset threshold, then the fourth number is the target number corresponding to the abnormal behavior characteristics of the channel. The fifth preset threshold is obtained by multiplying the target number release volume by a preset percentage. The target number release volume is the number of numbers released by the channel where the fourth number is located in the second preset time period.

[0167] The first preset threshold, the second preset threshold, the third preset threshold, the fourth preset threshold, and the preset percentage are determined based on the indicator data.

[0168] Here, each number to be identified can correspond to a target channel.

[0169] Here, the second preset condition is obtained by analyzing the indicator data of the sample number set. In one embodiment, based on new entrant numbers from the same network access channel, behavioral characteristic data such as call behavior, data service, and power on / off status within two weeks of network access are analyzed to obtain the features with the greatest distinguishing effect on channel anomaly rules. Cluster analysis is then used to obtain various preset thresholds, and the second preset condition is determined based on these preset thresholds.

[0170] It should be noted that the second preset condition includes at least one of the four channel anomaly rules: anomaly number rules with the same IMEI (International Mobile Equipment Identity), anomaly number rules with the same base station, anomaly number rules with the same top-ranked peer number, and anomaly number rules for calls made between devices within the same channel. The top-ranked peer number is the peer number with the most frequent communications.

[0171] Here, the second preset time period can be set according to actual needs, such as 1 week, 2 weeks, etc.

[0172] In one embodiment, the IMEI is an 11-bit code.

[0173] In one embodiment, the first peer number includes the peer number for a phone call and the peer number for a text message communication.

[0174] The card-raising identification method provided in this embodiment of the invention determines the target number corresponding to the abnormal channel behavior characteristics through a second preset condition, thereby increasing the identification of card-raising based on abnormal channel behavior characteristics and further improving the identification accuracy of card-raising.

[0175] Based on any of the above embodiments, in this method, the behavioral characteristics further include silent behavioral characteristics, and the target number corresponding to the silent behavioral characteristics is determined based on the following method:

[0176] Determine the outgoing call duration, incoming call duration, and data usage for each number in the set of numbers to be identified;

[0177] From the set of numbers to be identified, determine the fifth number whose call duration is less than or equal to a preset call duration threshold, whose call duration is less than or equal to a preset call duration threshold, and whose data usage is less than or equal to a preset data usage threshold, and use the fifth number as the target number corresponding to the silent behavior feature;

[0178] The preset outgoing call duration threshold, the preset incoming call duration threshold, and the preset traffic threshold are determined based on the indicator data.

[0179] Here, each number to be identified can correspond to a call duration for the calling party, a call duration for the receiving party, and data usage.

[0180] In one embodiment, the outgoing call duration, incoming call duration, and data usage of the identified SIM card farming numbers are analyzed. Based on the analysis results, preset outgoing call duration thresholds, preset incoming call duration thresholds, and preset data usage thresholds are determined. For example, the preset outgoing call duration threshold is 0, the preset incoming call duration threshold is 0, and the preset data usage threshold is 0.

[0181] The card nullification identification method provided in this invention determines the target number corresponding to the silent behavior characteristics by using the call duration of each number to be identified, the call duration of the incoming call, and the data usage. This can increase the identification of card nullification based on silent behavior characteristics, thereby further improving the identification accuracy of card nullification.

[0182] In any of the above embodiments, the behavioral characteristics in this method further include normal number behavioral characteristics, and the target number corresponding to the normal number behavioral characteristics is determined based on the following method:

[0183] Determine the average monthly recharge amount for each number in the set of numbers to be identified;

[0184] From the set of numbers to be identified, the sixth number whose average monthly recharge amount is greater than a preset amount threshold is determined, and the sixth number is used as the target number corresponding to the normal number behavior characteristics.

[0185] The preset amount threshold is determined based on the indicator data.

[0186] Here, each number to be identified can correspond to the average monthly recharge amount.

[0187] In one embodiment, cross-analysis is performed on the call behavior, data service usage, power on / off status, and average monthly consumption of known channel-manipulated phone numbers on the existing network, along with their average monthly recharge amount. A preset amount threshold is determined based on the analysis results. For example, the analysis reveals that among users with an average monthly recharge amount of less than 30 yuan, 68% of these numbers have fewer than two outgoing and incoming calls, 60% have less than 50MB of data usage, and 73% have been powered on for less than two days.

[0188] The card-farming identification method provided in this invention identifies normal numbers by their average monthly recharge amount, thereby identifying card-farming based on the behavioral characteristics of normal numbers, and further improving the accuracy of card-farming identification.

[0189] Based on any of the above embodiments, in this method, the target number includes abnormal numbers and numbers corresponding to normal number behavior characteristics. The abnormal numbers include numbers corresponding to the joint card farming group, numbers corresponding to the concentrated behavior characteristics of the counterpart number channel, numbers corresponding to the abnormal behavior characteristics of the channel, and numbers corresponding to the silent behavior characteristics.

[0190] The process of determining the card-maintenance number based on the target number includes:

[0191] Determine the union of the numbers corresponding to the joint card-raising group, the numbers corresponding to the channel-based concentrated behavior characteristics of the counterpart numbers, the numbers corresponding to the channel-based abnormal behavior characteristics, and the numbers corresponding to the silent behavior characteristics;

[0192] Based on the numbers corresponding to the normal number behavior characteristics, the normal numbers in the union set are removed to obtain the card-maintenance numbers.

[0193] Here, the numbers corresponding to normal number behavior characteristics can include one or more, the numbers corresponding to joint card farming groups can include one or more, the numbers corresponding to the concentrated behavior characteristics of the counterpart number channel can include one or more, the numbers corresponding to the abnormal behavior characteristics of the channel can include one or more, the numbers corresponding to the silent behavior characteristics can include one or more, and the card farming numbers can include one or more.

[0194] For ease of understanding, please refer to the following formula:

[0195] Card farming number = (number corresponding to the joint card farming group U, number corresponding to the channel's concentrated behavior characteristics U, number corresponding to the channel's abnormal behavior characteristics U, number corresponding to the silent behavior characteristics) - number corresponding to the normal number's behavior characteristics.

[0196] The card-farming identification method provided in this invention includes target numbers corresponding to abnormal and normal numbers based on their behavioral characteristics. This approach considers both abnormal and normal numbers, removing normal numbers from suspected card-farming numbers and preventing incorrect identification, thereby further improving the accuracy of card-farming identification. Furthermore, abnormal numbers include those corresponding to joint card-farming groups, numbers corresponding to concentrated behavioral characteristics of peer number channels, numbers corresponding to abnormal channel behavior characteristics, and numbers corresponding to silent behavior characteristics. This approach considers abnormal numbers (suspected card-farming numbers) from various dimensions, further improving the accuracy of card-farming identification.

[0197] The card maintenance identification device provided by the present invention is described below. The card maintenance identification device described below can be referred to in correspondence with the card maintenance identification method described above.

[0198] Figure 3 This is a schematic diagram of the card recognition device provided by the present invention, as shown below. Figure 3 As shown, the card recognition device includes:

[0199] The first determining module 310 is used to determine the target number corresponding to the behavioral characteristics from the set of numbers to be identified based on the indicator data of the sample number set;

[0200] The second determining module 320 is used to determine the card maintenance number based on the target number;

[0201] The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified;

[0202] The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm.

[0203] The card-farming identification device provided in this invention determines the target number corresponding to the behavioral characteristics from the set of numbers to be identified based on the indicator data of the sample number set; and determines the card-farming number based on the target number. Through this method, card-farming identification can be performed, and the card-farming number can be determined. The behavioral characteristics include joint card-farming behavioral characteristics. The joint card-farming group is obtained by comparing the similarity of the number trajectory of the set of numbers to be identified with the similarity threshold corresponding to the similarity of the number trajectory of the set of numbers to be identified, thereby determining the number corresponding to the joint card-farming group as the target number. Furthermore, the similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, indicator data, and the loss function of a greedy algorithm. Therefore, the similarity threshold can be automatically trained based on the number trajectory similarity, eliminating the need for manually defining card-farming identification rules and improving the accuracy and efficiency of card-farming identification.

[0204] Based on any of the above embodiments, the device further includes a threshold training module, which includes:

[0205] The threshold determination unit is used to determine the sample similarity threshold for the current iteration round;

[0206] The community determination unit is used to determine, based on the number trajectory similarity of the sample number set and the sample similarity threshold, a first number set corresponding to the joint card-raising community and a second number set corresponding to the non-joint card-raising community from the sample number set;

[0207] The threshold update unit is used to update the sample similarity threshold based on the index distance between the first number set and the second number set, and the loss function, to obtain the sample similarity threshold for the next iteration round, and to use the next iteration round as the current iteration round, until the current iteration round is the last iteration round, so as to obtain the similarity threshold. The index distance is determined based on the index data.

[0208] Based on any of the above embodiments, the indicator distance includes at least one of the distance between the number of calls made and received, the distance of data traffic, and the distance of the number of days the device is powered on.

[0209] Based on any of the above embodiments, the device further includes a group determination module, which includes:

[0210] A number determination unit is used to determine, from the set of numbers to be identified, the first number to be compared and the second number to be compared in the current comparison round;

[0211] The similarity determination unit is used to determine the number trajectory similarity between the first number to be compared and the second number to be compared based on the number of base stations of the first base station communicated by the first number to be compared within a first preset time period, the number of base stations of the second number to be compared communicated by the second number within the first preset time period, and the number of base stations of the common base stations communicated by the first number to be compared and the second number to be compared within the first preset time period. The number trajectory similarity between the first number to be compared and the second number to be compared includes at least one of the similarity between the common base station and the first base station, the similarity between the common base station and the second base station, and the number of days the common base station appears.

[0212] The similarity comparison unit is used to compare the number trajectory similarity of the first number to be compared and the second number to be compared with the similarity threshold to obtain a comparison result;

[0213] The step return unit is used to return the step of determining the first and second numbers to be compared in the current comparison round from the set of numbers to be identified, until the current comparison round is the last comparison round;

[0214] A group determination unit is used to determine the joint card-raising group based on the comparison results of the set of numbers to be identified.

[0215] Based on any of the above embodiments, the behavioral characteristics further include remote number channel-based behavioral characteristics, and the device further includes a number determination module, which is used for:

[0216] Determine the peer number of each peer number in the set of peer numbers to be identified, and the first target channel in which the peer number of each peer number to be identified is located;

[0217] The numbers to be analyzed in the current analysis round are determined from the sets of numbers to be identified. Based on the peer numbers of each number to be identified and the first target channel, the number of peer numbers of the number to be analyzed in each channel and the percentage of peer numbers of the number to be analyzed in each channel are determined.

[0218] When both the number of numbers and the proportion of numbers meet the first preset condition, the number to be analyzed is taken as the target number corresponding to the concentrated behavior characteristics of the peer number channel. The first preset condition is determined based on the indicator data.

[0219] Return to the step of determining the number to be analyzed in the current analysis round from each set of numbers to be identified, until the current analysis round is the last analysis round.

[0220] Based on any of the above embodiments, the behavioral characteristics further include channel abnormal behavior characteristics, and the device further includes a number determination module, which is further used for:

[0221] Determine the second target channel for each number in the set of numbers to be identified;

[0222] Based on the number to be identified, the second target channel, and the second preset conditions, the target number corresponding to the abnormal behavior characteristics of the channel is determined;

[0223] The second preset condition includes at least one of the following:

[0224] If the number of first numbers with the same International Mobile Equipment Identity (IMEI) on the same channel exceeds a first preset threshold, the first number is the target number corresponding to the channel's abnormal behavior characteristics.

[0225] If the number of second numbers on the same channel and at the same base station exceeds the second preset threshold, the second number is the target number corresponding to the abnormal behavior characteristics of the channel;

[0226] If the number of third numbers on the same channel and with the same first peer number is greater than a third preset threshold, the third number is the target number corresponding to the abnormal behavior characteristics of the channel, and the first peer number is the number that the third number communicates with the most.

[0227] If the number of fourth numbers that are on the same channel and have mutual calling behavior is greater than the fourth preset threshold and greater than the fifth preset threshold, then the fourth number is the target number corresponding to the abnormal behavior characteristics of the channel. The fifth preset threshold is obtained by multiplying the target number release volume by a preset percentage. The target number release volume is the number of numbers released by the channel where the fourth number is located in the second preset time period.

[0228] The first preset threshold, the second preset threshold, the third preset threshold, the fourth preset threshold, and the preset percentage are determined based on the indicator data.

[0229] Based on any of the above embodiments, the behavioral characteristics further include silent behavioral characteristics, and the device further includes a number determination module, which is further used for:

[0230] Determine the outgoing call duration, incoming call duration, and data usage for each number in the set of numbers to be identified;

[0231] From the set of numbers to be identified, determine the fifth number whose call duration is less than or equal to a preset call duration threshold, whose call duration is less than or equal to a preset call duration threshold, and whose data usage is less than or equal to a preset data usage threshold, and use the fifth number as the target number corresponding to the silent behavior feature;

[0232] The preset outgoing call duration threshold, the preset incoming call duration threshold, and the preset traffic threshold are determined based on the indicator data.

[0233] Based on any of the above embodiments, the behavioral characteristics further include normal number behavioral characteristics, and the device further includes a number determination module, which is further configured to:

[0234] Determine the average monthly recharge amount for each number in the set of numbers to be identified;

[0235] From the set of numbers to be identified, the sixth number whose average monthly recharge amount is greater than a preset amount threshold is determined, and the sixth number is used as the target number corresponding to the normal number behavior characteristics.

[0236] The preset amount threshold is determined based on the indicator data.

[0237] Based on any of the above embodiments, the target number includes numbers corresponding to abnormal numbers and numbers corresponding to normal number behavior characteristics. The abnormal numbers include numbers corresponding to the joint card-farming group, numbers corresponding to the concentrated behavior characteristics of the counterpart number channel, numbers corresponding to the abnormal behavior characteristics of the channel, and numbers corresponding to the silent behavior characteristics. The second determining module 320 includes:

[0238] The union determination unit is used to determine the union of the numbers corresponding to the joint card-raising group, the numbers corresponding to the channel-based concentrated behavior characteristics of the counterpart numbers, the numbers corresponding to the channel-based abnormal behavior characteristics, and the numbers corresponding to the silent behavior characteristics;

[0239] The card maintenance determination unit is used to obtain the card maintenance number by removing the normal numbers from the union set based on the numbers corresponding to the normal number behavior characteristics.

[0240] Based on any of the above embodiments, the indicator data includes at least one of the following: user basic information, call behavior, data services, base station information, consumption characteristics, social circles, channel characteristics, and power on / off characteristics.

[0241] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a card-farming identification method, which includes: determining the target number corresponding to the behavioral characteristics from the number set to be identified based on the indicator data of the sample number set; determining the card-farming number based on the target number; wherein the behavioral characteristics include joint card-farming behavioral characteristics, the target number corresponding to the joint card-farming behavioral characteristics is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with a similarity threshold corresponding to the number trajectory similarity of the number set to be identified; the similarity threshold is obtained by iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm.

[0242] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0243] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the card-raising identification method provided by the above methods. The method includes: determining a target number corresponding to a behavioral feature from a set of numbers to be identified based on indicator data of a sample number set; determining a card-raising number based on the target number; wherein the behavioral feature includes joint card-raising behavioral features, the target number corresponding to the joint card-raising behavioral features is a number corresponding to a joint card-raising group, and the joint card-raising group is obtained by comparing the number trajectory similarity of the set of numbers to be identified with a similarity threshold corresponding to the number trajectory similarity of the set of numbers to be identified; the similarity threshold is obtained by iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of a greedy algorithm.

[0244] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the card-raising identification method provided by the above methods. The method includes: determining a target number corresponding to a behavioral feature from a set of numbers to be identified based on indicator data of a sample number set; determining a card-raising number based on the target number; wherein the behavioral feature includes joint card-raising behavioral features, the target number corresponding to the joint card-raising behavioral features is a number corresponding to a joint card-raising group, and the joint card-raising group is obtained by comparing the number trajectory similarity of the set of numbers to be identified with a similarity threshold corresponding to the number trajectory similarity of the set of numbers to be identified; the similarity threshold is obtained by iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of a greedy algorithm.

[0245] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0246] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0247] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying prepaid cards, characterized in that, include: Based on the indicator data of the sample number set, the target number corresponding to the behavioral characteristics is determined from the number set to be identified; Based on the target number, determine the card-maintenance number; The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified; The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm; the indicator data of the sample number set is used to determine the distance between joint card-raising communities and non-joint card-raising communities; The similarity threshold is obtained through iterative training in the following manner: Determine the sample similarity threshold for the current iteration round; Based on the number trajectory similarity of the sample number set and the sample similarity threshold, a greedy algorithm is used to determine the first number set corresponding to the joint card-raising community and the second number set corresponding to the non-joint card-raising community from the sample number set; Based on the index distance between the first number set and the second number set, and using the reciprocal of the index distance as the loss function of the greedy algorithm, the sample similarity threshold is updated to obtain the sample similarity threshold for the next iteration round. The next iteration round is then used as the current iteration round until the current iteration round is the last iteration round, in order to obtain the similarity threshold. The index distance is determined based on the index data.

2. The card identification method according to claim 1, characterized in that, The distance indicator includes at least one of the following: distance between the number of calls made and received, distance of data traffic, and distance of the number of days the device is powered on.

3. The card identification method according to claim 1, characterized in that, The joint card-raising group was determined based on the following method: From the set of numbers to be identified, determine the first and second numbers to be compared in the current comparison round; Based on the number of base stations of the first base station that the first number to be compared communicated with during the first preset time period, the number of base stations of the second base station that the second number to be compared communicated with during the first preset time period, and the number of base stations of the common base stations that the first number to be compared and the second number to be compared communicated with during the first preset time period, the number trajectory similarity between the first number to be compared and the second number to be compared is determined. The number trajectory similarity between the first number to be compared and the second number to be compared includes at least one of the similarity between the common base station and the first base station, the similarity between the common base station and the second base station, and the number of days the common base station appears. The similarity of the number trajectories of the first number to be compared and the second number to be compared is compared with the similarity threshold to obtain the comparison result; Return to the step of determining the first and second numbers to be compared in the current comparison round from the set of numbers to be identified, until the current comparison round is the last comparison round; Based on the comparison results of the number set to be identified, the joint card-raising group is determined.

4. The card identification method according to claim 1, characterized in that, The behavioral characteristics also include the channel-based concentrated behavior characteristics of the peer number, and the target number corresponding to the channel-based concentrated behavior characteristics of the peer number is determined based on the following method: Determine the peer number of each peer number in the set of peer numbers to be identified, and the first target channel in which the peer number of each peer number to be identified is located; The numbers to be analyzed in the current analysis round are determined from the sets of numbers to be identified. Based on the peer numbers of each number to be identified and the first target channel, the number of peer numbers of the number to be analyzed in each channel and the percentage of peer numbers of the number to be analyzed in each channel are determined. When both the number of numbers and the proportion of numbers meet the first preset condition, the number to be analyzed is taken as the target number corresponding to the concentrated behavior characteristics of the peer number channel. The first preset condition is determined based on the indicator data. Return to the step of determining the number to be analyzed in the current analysis round from each set of numbers to be identified, until the current analysis round is the last analysis round.

5. The card identification method according to claim 1, characterized in that, The behavioral characteristics also include abnormal channel behavior characteristics, and the target number corresponding to the abnormal channel behavior characteristics is determined based on the following method: Determine the second target channel for each number in the set of numbers to be identified; Based on the number to be identified, the second target channel, and the second preset conditions, the target number corresponding to the abnormal behavior characteristics of the channel is determined; The second preset condition includes at least one of the following: If the number of first numbers with the same International Mobile Equipment Identity (IMEI) on the same channel exceeds a first preset threshold, the first number is the target number corresponding to the channel's abnormal behavior characteristics. If the number of second numbers on the same channel and at the same base station exceeds the second preset threshold, the second number is the target number corresponding to the abnormal behavior characteristics of the channel; If the number of third numbers on the same channel and with the same first peer number is greater than a third preset threshold, the third number is the target number corresponding to the abnormal behavior characteristics of the channel, and the first peer number is the number that the third number communicates with the most. If the number of fourth numbers that are on the same channel and have mutual calling behavior is greater than the fourth preset threshold and greater than the fifth preset threshold, then the fourth number is the target number corresponding to the abnormal behavior characteristics of the channel. The fifth preset threshold is obtained by multiplying the target number release volume by a preset percentage. The target number release volume is the number of numbers released by the channel where the fourth number is located in the second preset time period. The first preset threshold, the second preset threshold, the third preset threshold, the fourth preset threshold, and the preset percentage are determined based on the indicator data.

6. The card identification method according to claim 1, characterized in that, The behavioral characteristics also include silent behavior characteristics, and the target number corresponding to the silent behavior characteristics is determined based on the following method: Determine the outgoing call duration, incoming call duration, and data usage for each number in the set of numbers to be identified; From the set of numbers to be identified, determine the fifth number whose call duration is less than or equal to a preset call duration threshold, whose call duration is less than or equal to a preset call duration threshold, and whose data usage is less than or equal to a preset data usage threshold, and use the fifth number as the target number corresponding to the silent behavior feature; The preset outgoing call duration threshold, the preset incoming call duration threshold, and the preset traffic threshold are determined based on the indicator data.

7. The card identification method according to claim 1, characterized in that, The behavioral characteristics also include normal number behavioral characteristics, and the target number corresponding to the normal number behavioral characteristics is determined based on the following method: Determine the average monthly recharge amount for each number in the set of numbers to be identified; From the set of numbers to be identified, the sixth number whose average monthly recharge amount is greater than a preset amount threshold is determined, and the sixth number is used as the target number corresponding to the normal number behavior characteristics. The preset amount threshold is determined based on the indicator data.

8. The card identification method according to claim 1, characterized in that, The target number includes numbers corresponding to abnormal numbers and numbers corresponding to normal number behavior characteristics. The abnormal numbers include numbers corresponding to the joint card farming group, numbers corresponding to the concentrated behavior characteristics of the counterpart number channel, numbers corresponding to the abnormal behavior characteristics of the channel, and numbers corresponding to the silent behavior characteristics. The process of determining the card-maintenance number based on the target number includes: Determine the union of the numbers corresponding to the joint card-raising group, the numbers corresponding to the channel-based concentrated behavior characteristics of the counterpart numbers, the numbers corresponding to the channel-based abnormal behavior characteristics, and the numbers corresponding to the silent behavior characteristics; Based on the numbers corresponding to the normal number behavior characteristics, the normal numbers in the union set are removed to obtain the card-maintenance numbers.

9. The card identification method according to any one of claims 1 to 8, characterized in that, The indicator data includes at least one of the following: user basic information, call behavior, data services, base station information, consumption characteristics, social circles, channel characteristics, and power on / off characteristics.

10. A card recognition device, characterized in that, include: The first determination module is used to determine the target number corresponding to the behavioral characteristics from the set of numbers to be identified based on the indicator data of the sample number set; The second determining module is used to determine the card maintenance number based on the target number; The behavioral features include joint card-farming behavior features, the target number corresponding to the joint card-farming behavior features is the number corresponding to the joint card-farming group, and the joint card-farming group is obtained by comparing the number trajectory similarity of the number set to be identified with the similarity threshold corresponding to the number trajectory similarity of the number set to be identified; The similarity threshold is obtained through iterative training based on the number trajectory similarity of the sample number set, the indicator data, and the loss function of the greedy algorithm; the indicator data of the sample number set is used to determine the distance between joint card-raising communities and non-joint card-raising communities; The similarity threshold is obtained through iterative training in the following manner: Determine the sample similarity threshold for the current iteration round; Based on the number trajectory similarity of the sample number set and the sample similarity threshold, a greedy algorithm is used to determine the first number set corresponding to the joint card-raising community and the second number set corresponding to the non-joint card-raising community from the sample number set; Based on the index distance between the first number set and the second number set, and using the reciprocal of the index distance as the loss function of the greedy algorithm, the sample similarity threshold is updated to obtain the sample similarity threshold for the next iteration round. The next iteration round is then used as the current iteration round until the current iteration round is the last iteration round, in order to obtain the similarity threshold. The index distance is determined based on the index data.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the card identification method as described in any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the card identification method as described in any one of claims 1 to 9.

Images