Family member identification method and device, equipment, storage medium and product
The method enhances family member identification by analyzing traffic and location data to classify Internet behaviors and construct multi-dimensional features, accurately identifying family members across different networks using advanced models.
Patent Information
- Application Number
- CN202411187382.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-06
AI Technical Summary
Existing family member identification solutions are unable to identify family members across different networks due to the inability of a single operator to directly obtain location information of users on other networks.
A method that analyzes communication data, including traffic and location information, to identify family members by classifying Internet behaviors and constructing multi-dimensional features, using models like generative adversarial networks and decision trees to determine family relationships.
Improves the accuracy of family member identification by leveraging gateway traffic data to recognize family members across different networks, overcoming the limitations of single-operator data access.
Smart Images

Figure CN119095099A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of big data technology, and in particular to a family member identification method, device, equipment, storage medium and product. Background Art
[0002] With the rapid development of the four major telecom operators, both personal and family services are close to saturation and entering the stock market stage. In the face of the current situation, developing integrated services is the inevitable direction for seeking new business growth points, and the identification of family member relationships is an important link for integrating family services and personal services, and is an important data basis for the operation of personal and family integrated services.
[0003] Most existing family member identification solutions rely on traffic data for identification and are unable to identify family members on different networks. Summary of the invention
[0004] The main purpose of this application is to provide a family member identification method, device, equipment, storage medium and product, aiming to solve the technical problem of the existing solution that family members in different networks cannot be identified.
[0005] To achieve the above purpose, the present application proposes a method for identifying family members, the method comprising:
[0006] Acquire the collected communication data of the user number; the communication data includes the flow data generated by the user number when passing through each gateway;
[0007] Analyzing the traffic data to obtain an Internet behavior classification result, and determining whether the traffic data meets a preset home category condition based on the Internet behavior classification result;
[0008] It is determined that users corresponding to user numbers whose corresponding traffic data in the same gateway meet the preset home category conditions are members of the same family circle.
[0009] In one embodiment, the online behavior classification result includes the proportion of different types of online behaviors, and the steps of analyzing the traffic data to obtain the online behavior classification result, and determining whether the traffic data meets the preset home category condition based on the online behavior classification result include:
[0010] Inputting the traffic data into a pre-trained traffic classification model to obtain the proportion of different types of online behaviors;
[0011] Based on the proportions of the different categories of online behaviors, it is determined whether the traffic data meets a preset home category condition.
[0012] In one embodiment, the step of determining whether the traffic data meets a preset household category condition based on the proportion of the different categories of online behaviors includes:
[0013] Determining an online behavior score based on the proportions of the different types of online behaviors and predetermined weights of the different types of online behaviors;
[0014] Based on the comparison result between the online behavior score and a preset threshold, it is determined whether the traffic data meets a preset home category condition.
[0015] In one embodiment, the communication data further includes user information, signaling data, and location information corresponding to the user number, and the method further includes, after the step of acquiring the collected communication data of the user number:
[0016] Constructing a first multi-dimensional feature of each of the user numbers based on the user information, signaling data, and location information corresponding to each of the user numbers;
[0017] A plurality of numbers having potential family relationships are identified through the first multi-dimensional features of each of the user numbers.
[0018] In one embodiment, the location information includes a resident location data set, and before the step of identifying a plurality of user numbers having a potential family relationship through the first multi-dimensional features of each of the user numbers, the step further includes:
[0019] Calculating the similarity of the permanent location data sets between the user numbers to obtain the permanent location similarity between the user numbers;
[0020] The step of identifying a plurality of user numbers having a potential family relationship by using the first multi-dimensional feature of each of the user numbers comprises:
[0021] Based on the first multi-dimensional features of each of the user numbers and the similarity of the permanent locations between the user numbers, construct a second multi-dimensional feature of the number pair; the number pair is two different user numbers;
[0022] The second multi-dimensional features of the number pairs are input into a pre-trained social circle recognition model to determine whether there is a potential family relationship between the number pairs.
[0023] In one embodiment, the step of calculating the similarity of the permanent location data sets between the user numbers to obtain the permanent location similarity between the user numbers comprises:
[0024] Mapping the resident location data set of each user number to a preset inner product space based on a preset kernel function to obtain the data point to which each user number is mapped;
[0025] In the inner product space, the inner products between the data points to which the user numbers are mapped are calculated to obtain the resident position similarities between the user numbers.
[0026] In one embodiment, after the step of identifying a plurality of user numbers having a potential family relationship by using the first multi-dimensional features of each of the user numbers, the method further comprises:
[0027] Clustering user numbers whose corresponding traffic data in the same gateway meet preset home category conditions to obtain a first initial home circle;
[0028] Clustering multiple user numbers with potential family relationships to obtain a second initial family circle;
[0029] The first initial family circle and the second initial family circle are merged to obtain a merged family circle.
[0030] In one embodiment, the step of merging the first initial family circle and the second initial family circle to obtain a merged family circle includes:
[0031] Merging the user numbers in the first initial family circle and the second initial family circle to obtain a candidate family circle;
[0032] A seed number and a pairing number corresponding to the seed number are selected from the candidate family circle, and the family closeness between the seed number and the pairing number is determined; based on the family closeness, the corresponding pairing number is added to the family circle of the seed number to obtain the integrated family circle; wherein the seed number and the pairing number are both user numbers in the integrated family circle.
[0033] In one embodiment, the step of selecting a seed number and a pairing number corresponding to the seed number from the candidate family circle, determining the family closeness between the seed number and the pairing number, and adding the corresponding pairing number to the family circle of the seed number based on the family closeness to obtain the integrated family circle includes:
[0034] Selecting a seed number corresponding to an initial number of iterations from the candidate family circle, and initializing the family circle to be expanded of the seed number;
[0035] The family circle to be expanded of the seed number is subjected to multiple iterative processing, and each iterative processing includes: selecting each paired number corresponding to the seed number in the candidate family circle, and processing the communication data of the seed number and the communication data of each paired number according to a preset key feature extraction rule to obtain the key features between the seed number and each paired number; determining the family closeness between the seed number and each paired number based on the key features between the seed number and each paired number; adding the paired number whose family closeness is greater than a preset closeness threshold to the family circle to be expanded of the seed number, and selecting the paired number added to the family circle to be expanded as the seed number corresponding to the next number of iterations;
[0036] When it is detected that the number of number members in the family circle to be expanded of the seed number is greater than or equal to a preset limit threshold, the iterative processing is stopped, and the family circle to be expanded of the seed number when the iterative processing is stopped is determined, and the merged family circle is determined based on the family circle to be expanded of the seed number when the iterative processing is stopped.
[0037] In one embodiment, after the step of fusing the first initial family circle and the second initial family circle to obtain a merged family circle, the method further comprises:
[0038] Extracting features from the communication data of the user number in the integrated family circle according to a preset statistical feature extraction rule to obtain a plurality of statistical features;
[0039] Performing a weighted sum operation on the multiple statistical features to obtain a score;
[0040] Based on the interval to which the score belongs, the role of the user corresponding to the user number in the family is determined.
[0041] In one embodiment, the communication data further includes service information of the user number, and the method further includes, after the step of acquiring the collected communication data of the user number:
[0042] Determining service characteristics of each of the numbers based on service information of each of the numbers;
[0043] Based on the service characteristics of each of the numbers, multiple numbers with potential family relationships are identified through preset service rules to obtain a third initial family circle;
[0044] After the step of determining that the user corresponding to the user number whose corresponding traffic data in the same gateway meets the preset home category condition is a member of the same family circle, the method further includes:
[0045] Clustering user numbers whose corresponding traffic data in the same gateway meet preset home category conditions to obtain a first initial home circle;
[0046] The first initial family circle and the third initial family circle are merged to obtain a merged family circle.
[0047] In one embodiment, the step of merging the first initial family circle and the third initial family circle to obtain a merged family circle includes:
[0048] Merging corresponding numbers in the first initial family circle and the third initial family circle to obtain a candidate family circle;
[0049] A seed number and a pairing number corresponding to the seed number are selected from the candidate family circle, and the family closeness between the seed number and the pairing number is determined; based on the family closeness, the corresponding pairing number is added to the family circle of the seed number to obtain the integrated family circle, wherein the seed number and the pairing number are both user numbers in the integrated family circle.
[0050] In addition, to achieve the above-mentioned purpose, the present application also proposes a family member identification device, the family member identification device comprising:
[0051] A data acquisition module, used to acquire the collected communication data of the user number; the communication data includes the flow data generated by the user number when passing through each gateway;
[0052] A classification module, used to analyze the traffic data to obtain an Internet behavior classification result, and determine whether the traffic data meets a preset home category condition based on the Internet behavior classification result;
[0053] The determination module is used to determine that the users corresponding to the user numbers whose corresponding traffic data in the same gateway meet the preset home category conditions are members of the same family circle.
[0054] In addition, to achieve the above-mentioned purpose, the present application also proposes a family member identification device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the family member identification method described above.
[0055] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the family member identification method described above are implemented.
[0056] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the family member identification method described above.
[0057] The family member identification method, device, equipment, storage medium and product proposed in the present application obtain the collected communication data of the user number; the communication data includes the flow data generated by the user number when passing through each gateway; the flow data is analyzed to obtain the Internet behavior classification result, and based on the Internet behavior classification result, it is determined whether the flow data meets the preset home category condition; it is determined that the user corresponding to the user number whose corresponding flow data meets the preset home category condition in the same gateway is a member of the same family circle. The family member relationship is identified through the gateway flow data. Since the user equipment of the other network will connect to the gateway of this network for wireless Internet access, the flow data collected by the gateway includes not only the flow data of the users of this network, but also the flow data of the users of the other network. This can overcome the defect that a single operator cannot directly obtain the location information of the users of the other network, which leads to the inability to identify the family members of the other network, and improve the accuracy of family member identification. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0060] Figure 1 A schematic diagram of a process flow provided for the first embodiment of the method for identifying family members of the present application;
[0061] Figure 2 This is a flow chart of step S20 in the first embodiment of the method for identifying family members of the present application;
[0062] Figure 3 This is an application scenario diagram of the traffic classification model training based on the generative adversarial network involved in this application;
[0063] Figure 4 This is a schematic diagram of the network structure of the generator involved in this application;
[0064] Figure 5 This is a schematic diagram of the network structure of the identifier involved in this application;
[0065] Figure 6 A schematic diagram of a flow chart provided for the second embodiment of the method for identifying family members of the present application;
[0066] Figure 7 This is a flow chart of step S50 in Embodiment 2 of the method for identifying family members of this application;
[0067] Figure 8 This is a flow chart of the third embodiment of the method for identifying family members of the present application;
[0068] Fig. 9 A flowchart of the fourth embodiment of the method for identifying family members of the present application is provided;
[0069] Fig.10 This is a schematic diagram of the module structure of a family member identification device according to an embodiment of the present application;
[0070] Fig.11 This is a schematic diagram of the device structure of the hardware operating environment involved in the family member identification method in the embodiment of the present application.
[0071] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0072] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.
[0073] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.
[0074] The main solution of the embodiment of the present application is: by acquiring the communication data of the collected user number; the communication data includes the traffic data generated by the user number when passing through each gateway; the traffic data is analyzed to obtain the Internet behavior classification result, and based on the Internet behavior classification result, it is determined whether the traffic data meets the preset home category conditions; it is determined that the users corresponding to the user numbers whose corresponding traffic data in the same gateway meets the preset home category conditions are members of the same family circle.
[0075] Because the existing technologies mostly rely on traffic data or combine traffic data and location information for identification, and a single operator cannot directly obtain the location information of users on different networks, it is impossible to identify family members on different networks.
[0076] The present application provides a solution for identifying the relationship between family members through gateway traffic data. Since the user devices of different networks will connect to the gateway of this network for wireless Internet access, the traffic data collected by the gateway not only includes the traffic data of users of this network, but also includes the traffic data of users of different networks. This can overcome the defect that a single operator cannot directly obtain the location information of users of different networks, thereby making it impossible to identify family members of different networks, thereby improving the accuracy of family member identification.
[0077] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a family member identification device, etc. The following takes the family member identification device as an example to illustrate this embodiment and the following embodiments.
[0078] Based on this, the present application embodiment provides a method for identifying family members. Figure 1 , Figure 1 This is a flow chart of the first embodiment of the family member identification method of the present application.
[0079] In this embodiment, the family member identification method includes steps S10 to S30:
[0080] Step S10, acquiring the collected communication data of the user number; the communication data includes the flow data generated when the user number passes through each gateway.
[0081] In the present embodiment, the user number refers to the number in the user identity card (such as a SIM card). The communication data may include the flow data generated when the user number passes through each gateway within a preset first time period. Among them, there are multiple gateways, and each user number will generate corresponding flow data when passing through each gateway. For example, when the terminal device corresponding to user number 1 accesses the Internet through the home broadband gateway S, the corresponding flow data A_S1 is generated, when the terminal device corresponding to user number 2 accesses the Internet through the home broadband gateway S, the corresponding flow data A_S2 is generated, and when the terminal device corresponding to user number 1 accesses the Internet through the gateway D, the corresponding flow data A_D1 is generated. The flow data may include multiple data such as the Internet access address and the usage record of the application.
[0082] In addition, the communication data may also include user information, device information, service information, location information and signaling data corresponding to the user number collected within the preset first time period. User information may include the user's personal information, such as ID number, gender and age; service information may include service information such as home broadband service and family card service; signaling data may include call records, text message records, etc.
[0083] It should be noted that traffic data includes traffic data of mobile networks and fixed networks. In the use scenario of fixed networks, such as home broadband, there are usually multiple terminal devices connected to the same home broadband network. Among the multiple terminal devices, there may be devices using different operators. At this time, the gateway traffic data detected by the home broadband gateway includes not only the traffic data of the user numbers of this network, but also the traffic data of the user numbers of different networks. For example, mobile terminal 1 using the user identity card of operator A and mobile terminal 2 using the user identity card of operator B are both connected to the home broadband S of operator A and surf the Internet through the home broadband S. At this time, when collecting the gateway traffic data of the home broadband S, the user traffic data of operators A and B can be collected.
[0084] Table 1 below gives an example of some fields of the collected communication data:
[0085] Table 1
[0086] Field Tags type Field Chinese Description phone_number int User Number gender Varchar2 gender age Char age join_month Char Network access time level Varchar2 Star Rating ...... ...... ...... lag3_month_cnt float Number of calls in the past n months lag3_month_flux float Traffic usage in the past n months
[0087] Step S20, analyzing the traffic data to obtain an Internet behavior classification result, and determining whether the traffic data meets a preset home category condition based on the Internet behavior classification result.
[0088] In the present embodiment, the traffic data is the traffic data generated when any user number passes through any gateway. Regardless of whether it is a user of the local network or a user of another network, corresponding traffic data will be generated when accessing the gateway of the local operator and surfing the Internet through the gateway. Therefore, for any gateway, the traffic data passing through the gateway may be the traffic data of the user number of the local network or the traffic data of the user number of another network. Therefore, the family members of the other network can be mined through the gateway traffic data.
[0089] In this embodiment, the flow data is analyzed to identify the online behavior classification results corresponding to the flow data. As an example, the online behavior may include multiple categories such as gaming, learning and education, online shopping, online meetings, web browsing, and data downloading.
[0090] As an example, when the Internet behavior classification result includes the proportion of Internet behaviors of different categories, the Internet behavior score can be determined according to the proportion of Internet behaviors of different categories and the pre-set weights corresponding to Internet behaviors of different categories. When the Internet behavior score is greater than the preset threshold, it is determined that the traffic data meets the preset home category condition; otherwise, it is determined that the traffic data does not meet the preset home category condition. Among them, the weights corresponding to different categories of Internet behaviors are set according to actual conditions. For example, for Internet behaviors of the learning and education category, they are more inclined to be used in home places, so the category is set to have a higher weight. Alternatively, the first-level application category corresponding to each Internet behavior category can be determined based on preset rules, and each Internet behavior category corresponds to a first-level application category. The first-level application category can be a home category or an office category, and then it is identified whether the first-level application category with the largest proportion belongs to the home category. If it belongs to the home category, it is determined that the traffic data meets the preset home category condition, otherwise it is determined that the traffic data does not meet the preset home category condition.
[0091] In a feasible implementation, the above step S20 may include: inputting the traffic data into a pre-trained Internet behavior classification model to determine whether the main Internet behavior category corresponding to the traffic data belongs to the home category. When the main Internet behavior category belongs to the home category, it is determined that the traffic data meets the preset home category condition, otherwise it is determined that the traffic data does not meet the preset home category condition. The Internet behavior classification model is used to identify whether the main Internet behavior category corresponding to the input traffic data belongs to the preset home category. That is, the Internet behavior classification model is a binary classification model, and the classification results include belonging to the home category and not belonging to the home category.
[0092] In another feasible implementation, the above step S20 may include: inputting the traffic data into a pre-trained traffic classification model, determining the proportion of different categories of Internet behaviors corresponding to the traffic data, and determining whether the traffic data meets the preset home category conditions based on the proportion of different categories of Internet behaviors.
[0093] Compared with the first implementation, the model in this implementation classifies the proportion of different types of online behavior features, and the classified online behaviors are more detailed. The weights of online behavior categories can be flexibly set according to actual business characteristics, and the flexibility is strong.
[0094] Step S30, determining that users corresponding to user numbers whose corresponding traffic data in the same gateway meet preset home category conditions are members of the same family circle.
[0095] In the present embodiment, when the traffic data corresponding to any user number at any gateway meets the preset home category condition, it means that the traffic data generated by the user corresponding to the user number when surfing the Internet through the gateway is mainly used for home. Therefore, the users corresponding to the numbers whose corresponding traffic data in the same gateway meet the preset home category condition can be determined as members of a family circle. For example, assuming that the traffic data corresponding to user number 1 and user number 2 at gateway 1 meet the preset home category condition, it can be determined that the user corresponding to user number 1 and the user corresponding to user number 2 are likely to have a family relationship, and user number 1 and user number 2 are clustered into the same family circle.
[0096] In the solution of this embodiment, each gateway can identify at most one corresponding family circle, and thus multiple family circles can be identified by analyzing the traffic data of multiple gateways.
[0097] This embodiment scheme acquires the communication data of the collected user number; the communication data includes the flow data generated when the user number passes through each gateway; the flow data is analyzed to obtain the Internet behavior classification result, and based on the Internet behavior classification result, it is determined whether the flow data presets the home category condition; it is determined that the user corresponding to the user number whose corresponding flow data in the same gateway meets the preset home category condition is a member of the same family circle. The family member relationship is identified through the gateway flow data. Since the user equipment of the other network will connect to the gateway of this network for wireless Internet access, the flow data collected by the gateway includes not only the flow data of the users of this network, but also the flow data of the users of the other network. This can overcome the defect that a single operator cannot directly obtain the location information of the users of the other network, which leads to the inability to identify the family members of the other network, and improve the accuracy of family member identification.
[0098] Based on the above first embodiment, the above step S10 may further include: cleaning and standardizing the collected communication data. When cleaning the communication data, the communication data may be subjected to outlier removal, null value processing, deduplication, error format correction, and date and time format standardization. Among them, outlier removal may include removing abnormal user numbers, such as fraudulent users, suspected card-maintaining users, non-online user numbers, dismantling user numbers, public toll-free user numbers, wireless business calls, non-personal account opening certificates, electronic student ID user numbers, Internet of Things card users, and other abnormal user numbers.
[0099] By cleaning and standardizing communication data, the quality and consistency of the data can be ensured, thereby improving the accuracy and reliability of subsequent data processing and analysis.
[0100] Based on the first embodiment above, see Figure 2, the above step S20 may include steps S201 to S202:
[0101] Step S201, inputting the traffic data into a pre-trained traffic classification model to obtain the proportion of different types of Internet access behaviors.
[0102] In this embodiment, the traffic classification model is used to determine the proportion of different types of Internet access behaviors corresponding to the corresponding user number at the corresponding gateway (e.g., gateway 1) based on traffic data. The traffic classification model can be learned based on a neural network or other deep learning model, which is not limited in this application.
[0103] As an example, the above-mentioned traffic classification model can be generated based on a generative adversarial network, thereby reducing the workload of sample data labeling.
[0104] Specifically, see Figure 3 , Figure 3 The application scenario diagram of the traffic classification model training is shown. The above traffic classification model is obtained through the following steps e1 to e2 training:
[0105] Step e1, obtaining real traffic data samples. The real traffic data samples include marked data samples and unmarked data samples, and the samples include traffic data packets of user numbers at corresponding gateways. Then, each data sample is filtered to filter redundant data, for example, using any one of the Address Resolution Protocol (ARP), Dynamic Host Configuration Protocol (DHCP) or Internet Control Message Protocol (ICMP) to filter, and then normalized into a packet vector.
[0106] Step e2, cleaning, deduplication and standardization of the real traffic data samples after redundancy removal and normalization, and then inputting the standardized real traffic data samples into the traffic classification model based on the generative adversarial network for training.
[0107] In this embodiment, a combination of supervised and unsupervised methods is used when training the traffic classification model, which can significantly improve the learning ability of the model and effectively solve the problem that the traffic data of each application needs to be trained separately. In addition, semi-supervised learning can also overcome the limitations of labeled data.
[0108] Specifically, the generative adversarial network consists of two parts: the generator G and the discriminator D. The role of the generator G is to take random noise as input by learning the feature distribution of real data. The role of the discriminator D is to determine whether the data is real or generated by the generator G. Specifically, in the semi-supervised training process, the discriminator D and the generator G are trained alternately, and the parameters are updated iteratively. On the one hand, when training the generator, feature learning will be used to make the generated traffic as close to the real data distribution as possible. On the other hand, when training the discriminator, the network parameters are iteratively updated to minimize the cross entropy of the labeled traffic samples and the predicted probability distribution. Through adversarial training, the network model parameters of unlabeled samples and generated samples are updated. The model input is a data sample, and the output is an N+1-dimensional vector representing the class probability.
[0109] See also Figure 4 , Figure 4 The network structure diagram of the generator is given. The generator G consists of a deconvolution network and a convolution network. First, a 100-dimensional random noise z with Gaussian distribution is input into the fully connected network and converted into a three-dimensional vector through dimension transformation. The vector after dimension transformation is input into a convolution kernel w with a size of 4*4 and a stride of 2. It is activated by the activation function and then input into a convolution layer z with a kernel size of 7*7 and a stride of 2. Next, it is activated using the Tanh function to generate a 20*74*1 fake traffic sample.
[0110] See also Figure 5 , Figure 5 The network structure diagram of the discriminator is given. The discriminator D consists of 3 convolutional layers and 2 fully connected layers. After preprocessing the input data, the real traffic sample with a length of 1480 is converted into a three-dimensional vector (20*74*1), and then sent to a 3-layer convolution kernel w with a size of 3*3. LeakyReLU is used for activation after each convolution. LeakyReLU can retain a very small slope on the negative semi-axis (this scheme sets it to 0.2, which can be adjusted according to the model effect). After passing through the flattening layer, the fully connected layer is input. The network is constructed as a stacked network with shared weights, using Lambda as the activation function, and then outputs the probability of the normalized category.
[0111] Specifically, the model training steps include steps b1 to b5:
[0112] Step b1, generate a random vector with Gaussian noise z, and input it into the generator G to generate a false traffic data sample G(z);
[0113] Step b2, input the generated false traffic data samples G(z), the marked real traffic data samples and the unmarked real traffic data samples x into the discriminator network D in batches, and then output the sum of the probabilities of the classification results D(x) of the real traffic data samples and the classification results D[G(z)] of the false traffic data samples through the activation function;
[0114] Step b3, update the parameters of the generator network G. Use the gradient descent method to adjust the parameters of the discriminator D;
[0115] Step b4, fix the parameters of the discriminator network D, perform feature matching operations on the real traffic data sample x and the generated traffic G(z), then select the output of the fully connected layer as the middle layer, and use feature matching to adjust the parameters of the network G;
[0116] Step b5, repeating steps b1 to b4 until a preset iteration stop condition is reached, for example, a preset number of iterations is reached.
[0117] The above is an example process of the traffic classification model training given in this embodiment. Since it is a prior art, it will not be described in detail here.
[0118] Step S202: determining whether the traffic data meets a preset household category condition based on the proportion of different categories of online behaviors.
[0119] In a feasible implementation manner, the above step S202 may include:
[0120] Based on the proportions of different online behaviors and the predetermined weights of different categories of online behaviors, an online behavior score is determined; then, based on the comparison result of the online behavior score with a preset score threshold, it is determined whether the traffic data meets the preset home category condition.
[0121] In this embodiment, the sum of the product of the proportion of each category of online behavior and the weight can be calculated to obtain the online behavior score. For example, assuming that the proportions of learning and education, online shopping, and online meetings are a1%, a2%, and a3%, respectively, and assuming that learning and education, online shopping, and online meetings are further divided into home category, home category, and office category, respectively, and assuming that the weights of home category and office category are 80 and 20, respectively, then the online behavior score can be obtained by calculating a1%*80+a2%*80+a3%*20. Then, when the online behavior score is greater than the score threshold of the corresponding gateway, it is determined that the traffic data corresponding to the corresponding user number meets the preset home category condition. When the online behavior score is less than or equal to the score threshold of the corresponding gateway, it is determined that the traffic data of the corresponding user number does not meet the preset home category condition.
[0122] In another feasible implementation, the above step S202 may include: determining whether the total proportion of the Internet behavior categories belonging to the home category is greater than a preset proportion threshold, so as to determine whether the corresponding traffic data meets the preset home category condition. Specifically, when the total proportion is greater than the preset proportion threshold, it is determined that the corresponding traffic data meets the preset home category condition, otherwise it is determined that the corresponding traffic data does not meet the preset home category condition.
[0123] Through the above implementation plan, the proportion of different categories of user online behaviors is determined through the traffic classification model, and the proportion of different categories of online behaviors is analyzed to determine whether the traffic data meets the preset home category conditions. The weights of different categories of online behaviors can be flexibly adjusted in combination with actual scenarios, and the method is highly practical.
[0124] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction, and no further description will be given later. On this basis, the communication data also includes user information, signaling data and location information corresponding to the user number, please refer to Figure 6 After step S10, the family member identification method further includes steps S40 to S50:
[0125] Step S40: constructing a first multi-dimensional feature of each user number based on the user information, signaling data and location information corresponding to each user number.
[0126] In the present embodiment, feature extraction can be performed on user information of each user number based on a preset user feature extraction rule to obtain at least one feature of each user number in the user feature dimension, and feature extraction can be performed on signaling data of each user number based on a preset communication feature extraction rule to obtain at least one feature of each user number in the communication feature dimension, and feature extraction can be performed on location information of each user number based on a preset location feature extraction rule to obtain at least one feature of each user number in the location feature dimension, thereby obtaining the first multi-dimensional feature of each user number. The first multi-dimensional feature of each user number at least includes multiple features in three dimensions: user feature, communication feature, and location feature.
[0127] As an example, the extracted first multi-dimensional features may include the local user number, the opposite user number, the province code of the workplace, the cell code base station of the workplace, the cell name of the workplace, the number of time periods at the workplace, the number of days of residence at the workplace, the province code of the residence, the cell code base station of the residence, the cell name of the residence, the longitude of the base station of the residence, the latitude of the base station of the residence, the length of residence at the residence, the average number of time periods per month at the residence, the number of days of residence at the residence, the province code of the night permanent residence, the city code of the night permanent residence, the county code of the night permanent residence, the cell code base station of the night permanent residence, the cell name of the night permanent residence, the longitude of the night permanent base station, the latitude of the night permanent base station, the type of the night permanent base station, the type of the night permanent residence area, the average number of time periods per month at the night residence, the number of days of residence at the night permanent residence, the province code of the weekend permanent residence, the city code of the weekend permanent residence, the county code of the weekend permanent residence, the cell code base station of the weekend permanent residence, the cell name of the weekend permanent residence, the longitude of the weekend permanent base station, the latitude of the weekend permanent base station, and the type of the weekend permanent base station. type, weekend permanent residence area type, number of weekend permanent residence time periods, average number of weekend permanent residence time periods per month, number of weekend permanent residence days, weekend daytime permanent residence province code, weekend daytime permanent residence city code, weekend daytime permanent residence county code, weekend daytime permanent residence cell code base station, weekend daytime permanent residence cell name, weekend daytime permanent residence base station longitude, weekend daytime permanent residence base station latitude, weekend daytime permanent residence base station latitude, weekend daytime permanent residence base station type, weekend daytime permanent residence residence time, weekend daytime permanent residence The number of time periods in the area, the average number of time periods per month in the permanent area during the day on weekends, the number of days of residence in the permanent area during the day on weekends,..., the duration of callers' calls, the duration of callers' calls on weekdays, the duration of callees' calls on weekends, the duration of callees' calls, the duration of callees' calls on weekdays, the duration of callees' calls on weekends, the number of voice calls, the number of callers' calls on weekdays, the number of callers' calls on weekends, the number of callees' calls, the number of callees' calls on weekdays, the number of callees' calls on weekends, the number of callees' calls on weekdays, the number of callees' calls on weekends, the number of call days, the number of call days on weekends, the number of callers' calls and the number of callees' calls, etc.
[0128] Step S50: identifying multiple user numbers with potential family relationships through the first multi-dimensional features of each user number.
[0129] As a feasible implementation, the above step S50 may include: calculating the similarity of the permanent location data sets between each user number to obtain the permanent location similarity between each user number; constructing a second multidimensional feature of the number pair based on the first multidimensional feature of each user number and the permanent location similarity between each user number; the number pair is two different numbers; and identifying whether there is a potential family relationship between the number pair through the second multidimensional feature of the number pair.
[0130] As another feasible implementation, the above step S50 may include: calculating the correlation between each first feature between each user number through a preset correlation analysis algorithm, and determining whether there is a potential family relationship between each user number through the correlation between each first feature. Wherein, the first feature is a feature in the first multi-dimensional feature. Through the scheme of this embodiment, by combining user information, signaling data and location information for analysis, the family member identification relationship can be mined from multiple dimensions, thereby improving the accuracy of family member identification.
[0131] Based on the above second embodiment, please refer to Figure 7 , the above step S50 may include steps S501 to S503:
[0132] Step S501, calculating the similarity of the permanent location data sets between the various user numbers to obtain the permanent location similarity between the various user numbers.
[0133] In this embodiment, the resident location data set of the first user number and the resident location data set of the second user number in two different user numbers can be respectively estimated by kernel density, and the kernel density curve corresponding to the first user number and the kernel density curve corresponding to the second user number can be obtained, and the similarity between the resident location data set of the first user number and the resident location data set of the second user number can be obtained by calculating the ratio between the peak value of the kernel density curve corresponding to the first user number and the peak value of the kernel density curve corresponding to the second user number. Alternatively, the resident location data set of the first user number and the resident location data set of the second user number can be mapped to the inner product space by a preset kernel function, and the inner product of the data points corresponding to the first user number and the data points corresponding to the second user number can be calculated in the inner product space to obtain the resident location similarity of the user number pair.
[0134] Step S502: construct a second multi-dimensional feature of a number pair based on the first multi-dimensional feature of each user number and the similarity of the permanent positions between each user number; the number pair is two different user numbers.
[0135] In this embodiment, the first multidimensional feature of the first user number in the number pair, the first multidimensional feature of the second user number, and the resident position similarity between the first user number and the second user number can be concatenated to obtain the second multidimensional feature of the number pair.
[0136] Furthermore, the number pair may also be two user numbers that have a call relationship.
[0137] Step S503, inputting the second multi-dimensional features of the number pairs into a pre-trained social circle recognition model to determine whether there is a potential family relationship between the number pairs.
[0138] In this embodiment, the social circle model is used to identify whether there is a potential family relationship between number pairs.
[0139] As an example, the social circle recognition model can be obtained by constructing and training a decision tree. Specifically, when constructing and training the decision tree model, steps B11 to B13 may be included:
[0140] Step B11, obtain a training data set; the training data set includes a positive sample set and a negative sample set; the positive samples are the communication data of user number pairs that simultaneously meet the same account payment, same ID number card, and family card business relationships, and the negative samples are the communication data of number pairs that do not meet any family business relationship but have a call relationship.
[0141] Step B12, performing feature extraction on each training sample in the training data set, extracting multiple features in multiple dimensions including user features, communication features and location features, wherein the location features include resident location similarity.
[0142] Step B13, constructing a decision tree through the above multi-dimensional features, and training the decision tree through the training data set until the preset training stop condition is met.
[0143] Specifically, after obtaining the multi-dimensional features, the features with a high null value rate can be eliminated, and then the feature correlation coefficient can be calculated, and the features with a high correlation coefficient can be screened. After the feature variables are input into the decision tree model for training, the feature importance is output for feature selection, and the features with a feature importance of 0 are eliminated from the multi-dimensional features, and the eliminated multi-dimensional features are re-input into the decision tree for training to obtain the trained decision tree model. The trained decision tree model is used to determine whether there is a potential family relationship between the number pairs.
[0144] Since the method of establishing a decision tree is an existing technology, it will not be described in detail.
[0145] Furthermore, the above step S501 may include steps C1 to C2:
[0146] Step C1, mapping the resident location data set of each user number into a preset inner product space based on a preset kernel function, and obtaining the data point to which each user number is mapped.
[0147] It should be noted that the kernel function provides a method for mapping data into a high-dimensional space for processing, so the original low-dimensional resident location data set of each user number can be mapped into a high-dimensional space through the function. The inner product space is a vector space that defines an inner product operation, usually expressed as a dot product or projection between two vectors. The inner product space is usually used to define a kernel function so that the data in the original space can be mapped to a higher-dimensional space through the kernel function, thereby solving nonlinear problems. In this embodiment, the preset kernel function and inner product space can be selected based on actual conditions, and this application does not limit this.
[0148] As an example, the above step C1 may include steps C11 to C13:
[0149] Step C11, using a kernel function embedded with a kernel mean, processing the permanent location data set of the first user number and the permanent location data set of the second user number to obtain a distribution kernel between the permanent location data set of the first user number and the permanent location data set of the second user number:
[0150]
[0151] In formula (1), P X and P Y represents the distribution of the permanent location data set of the first user number and the distribution of the permanent location data set of the second user number, respectively, G (P X ,P Y ) represents the distribution kernel between the permanent location dataset of the first user number and the permanent location dataset of the second user number, and X and Y are respectively the distribution kernels of the permanent location datasets from the two distributions P X and P Y In the two sample groups generated by independent and identical distribution, x is a sample selected from sample group X, y is a sample selected from sample group Y, and k(x,y) is the isolation kernel between sample x and sample y.
[0152] Step C12, using the isolation kernel k to extract a finite-dimensional feature map from the permanent location data set of the first user number and the permanent location data set of the second user number Right now Transform formula (1) into:
[0153]
[0154] In formula (2), D represents the dataset of the permanent location dataset of the first user number and the permanent location dataset of the second user number; k(x,y|D) represents the isolation kernel between sample x and sample y under the given dataset D, K I (PX ,P Y | D) represents the distribution kernel between the distribution of the permanent location data set of the first user number and the distribution of the permanent location data set of the second user number under a given data set D, represents the mapping of the kernel mean k, and They represent the distribution P of the permanent location dataset of sample x, sample y, and the first user number under a given dataset D. X and the distribution P of the permanent location data set of the second user number Y The data points mapped by the kernel mean k are It represents the inner product between the data points mapped to sample x by kernel mean k and the data points mapped to sample y by kernel mean k under a given data set D. It represents the distribution P of the permanent location data set of the first user number under a given data set D. X The distribution P of the data points mapped by the kernel mean k and the permanent location data set of the second user number Y The inner product between the data points mapped by the kernel mean k.
[0155] Step C13, mapping the distribution X of the permanent location data set of the first user number to a point in the Hilbert space, and mapping the distribution Y of the permanent location data set of the second user number to a point in the Hilbert space.
[0156] Specifically, taking the distribution X of the permanent location data set of the first user number as an example, the mapping of the kernel mean k in the above formula (2) can be used: The distribution X of the permanent location data set of the first user number is mapped to a data point in the Hilbert space by the following formula (3):
[0157]
[0158] Step C2, calculating the inner product between the data points to which each user number is mapped in the inner product space, and obtaining the resident position similarity between each user number.
[0159] Since the calculation method of the inner product is an existing technology, it will not be described here.
[0160] Through this implementation, by constructing location features based on the similarity pairs of permanent locations of user number pairs, the correlation between the location features of two user numbers can be analyzed more deeply, and the potential family relationship between the user numbers can be accurately discovered.
[0161] Based on the above first and second embodiments, in the third embodiment of the present application, the same or similar contents as those of the above first and second embodiments can be referred to the above introduction, and will not be described in detail later. Figure 8 , after step S50, steps S60 to S80 may also be included:
[0162] Step S60: clustering the user numbers whose corresponding traffic data in the same gateway meet the preset home category condition to obtain a first initial home circle.
[0163] In this embodiment, user numbers whose corresponding traffic data in the same gateway meet preset home category conditions are clustered to obtain a first initial family circle. The first initial family circle may include one or more.
[0164] Step S70: cluster multiple user numbers with potential family relationships to obtain a second initial family circle.
[0165] In this embodiment, after multiple user numbers with potential family relationships are identified through the first multi-dimensional features of each user number, the multiple user numbers with potential family relationships are clustered into a family circle to obtain a second initial family circle.
[0166] Step S80: Merge the first initial family circle and the second initial family circle to obtain a merged family circle.
[0167] In a feasible implementation manner, the above step S80 may include: merging the user numbers in the first initial family circle and the second initial family circle according to a preset merging rule to obtain a merged family circle.
[0168] As an example, the preset merging rule may be: merging the user numbers in the two family circles with the highest overlap rate between the user numbers in the first initial family circle and the second initial family circle to obtain a merged family circle. For example, assuming that there are a first initial family circle A and a first initial family circle B, as well as a second initial family circle C and a second initial family circle D, and the number of identical mobile phone user numbers between the first initial family circle A and the second initial family circle D is greater than the number of identical mobile phone user numbers between the first initial family circle A and the other second initial family circles, it is determined that the overlap rate of mobile phone user numbers between A and the second initial family circle D is the highest, and the user numbers in the first initial family circle A and the second initial family circle D are merged, and the merged family circle contains all the user numbers of the first initial family circle A and the second initial family circle D.
[0169] As another example, the preset merging rule may be: merging two family circles in the first initial family circle with the same family decision maker as in the second initial family circle. For example, assuming that there are a first initial family circle A and a first initial family circle B, and a second initial family circle C and a second initial family circle D, and it is identified that the family decision makers of the first initial family circle A, the first initial family circle B, the second initial family circle C, and the second initial family circle D are mobile phone user number 1, mobile phone user number 2, mobile phone user number 1, and mobile phone user number 2, respectively, then the user numbers in the first initial family circle A and the second initial family circle C can be merged, and the user numbers in the first initial family circle B and the second initial family circle D can be merged.
[0170] Furthermore, when determining the family decision maker, the family decision makers of each family circle in the first initial family circle and the family decision makers of each family circle in the second initial family circle may be determined based on the service information of each user number and the preset decision maker identification rule. The preset decision maker identification rule may include: the corresponding user number is the handler of the family broadband service or the handler of the family card service.
[0171] In another feasible implementation, the above step S80 may include: merging the corresponding numbers in the first initial family circle and the second initial family circle to obtain a candidate family circle; selecting a seed number and a pairing number corresponding to the seed number in the candidate family circle, and determining the family closeness between the seed number and the pairing number based on the key features between the seed number and the pairing number, and adding the corresponding pairing number to the family circle of the seed number based on the family closeness to obtain a fused family circle. The key features between the seed number and the pairing number are obtained by processing the communication data of the seed number and the pairing number based on the preset key feature rules. Compared with the first implementation, this implementation can correct the relationship between family members based on the family closeness between the numbers, so that the accuracy of family member identification is higher.
[0172] Through this implementation scheme, by combining signaling data, location information and gateway traffic data to mine family member relationships, family members can be identified more comprehensively. At the same time, combining gateway traffic data to identify family member relationships can overcome the defect that a single operator cannot directly obtain the location information of users on different networks, which leads to the inability to identify family members on different networks, thereby improving the accuracy of family member identification and having strong applicability.
[0173] Based on the third embodiment, the step S80 may include steps d1 to d2:
[0174] Step d1: Merge corresponding user numbers in the first initial family circle and the second initial family circle to obtain a candidate family circle.
[0175] Specifically, the user numbers in the first initial family circle and the second initial family circle may be merged based on a preset merging rule to obtain a candidate family circle. The preset merging rule may refer to the above-mentioned related content and will not be described in detail here.
[0176] Step d2: Select a seed number and a pairing number corresponding to the seed number in the candidate family circle, determine the family closeness between the seed number and the pairing number, and add the corresponding pairing number to the family circle of the seed number based on the family closeness to obtain a merged family circle. The seed number and the pairing number are both user numbers in the candidate family circle.
[0177] In this embodiment, the above step d2 may include steps d21 to d23:
[0178] Step d21, select a seed number corresponding to the initial iteration number from the candidate family circles, and initialize the family circle to be expanded of the seed number.
[0179] Specifically, a number is randomly selected from each candidate family circle as a seed number corresponding to the initial iteration number, and then the family circle to be expanded of the seed number is initialized. The family circle to be expanded corresponding to the initialized seed number only contains the seed number.
[0180] Step d22, performing multiple iterative processing on the family circle to be expanded of the seed number, each iterative processing includes: selecting each pairing number corresponding to the seed number in the candidate family circle, and processing the communication data of the seed number and the communication data of each pairing number according to the preset key feature extraction rules to obtain the key features between the seed number and each pairing number; determining the family closeness between the seed number and each pairing number based on the key features between the seed number and each pairing number; adding the pairing numbers whose family closeness is greater than the preset closeness threshold to the family circle to be expanded of the seed number, and selecting the pairing numbers added to the family circle to be expanded as the seed number corresponding to the next number of iterations.
[0181] In the present embodiment, based on the communication data of each number, the communication data of each number can be feature extracted in combination with the preset key feature extraction rules, thereby constructing the key features between the seed number and each paired number. As an example, the preset key feature extraction rules may include: extracting the number of shared service relationships between the seed number and its paired number, the duration of the shared service relationship, the call frequency within a preset time period, the call month within a preset time period, the similarity of the permanent location, etc.
[0182] Step d23, when it is detected that the number of number members included in the family circle to be expanded of the seed number is greater than or equal to the preset limit threshold, the iterative processing of the seed number is stopped, and the family circle to be expanded of the seed number when the iterative processing is stopped is determined, and the integrated family circle is determined based on the family circle to be expanded of the seed number when the iterative processing is stopped.
[0183] Specifically, in the initial number of iterations, after the seed number is determined, other seed numbers other than the seed number are selected from the candidate family circle corresponding to the seed number as the pairing number, and then the family closeness between the seed number and each pairing number is calculated, and then the pairing numbers with family closeness greater than the preset closeness threshold can be added to the family circle to be expanded of the seed number in order of family closeness from large to small, until it is detected that the number of number members in the family circle to be expanded of the seed number is greater than or equal to the preset limit threshold. If in the initial number of iterations, the number of number members in the family circle to be expanded of the seed number does not reach the preset limit threshold, the pairing number added to the family circle to be expanded of the seed number can be selected as the seed number corresponding to the next number of iterations, and the pairing number with the largest family closeness can be selected randomly or selected. Afterwards, the iterative processing of the family circle to be expanded of the seed number is repeated until it is detected that the number of number members in the family circle to be expanded of the seed number is greater than or equal to the preset limit threshold. Among them, the preset limit threshold is set according to the actual situation.
[0184] Specifically, in each iteration, calculating the closeness between the seed number and the paired number may include steps f1 to f3:
[0185] Step f1, based on the key data information of each user number in the data sample set, determine the key features between the seed number and the paired number corresponding to each sliding window period.
[0186] Among them, the key data information may include multiple data such as business information, call signaling data and location information of each sliding window period, and then feature extraction is performed according to a preset key feature extraction rule to obtain the key features of each sliding window period. Each sliding window period is a plurality of window periods divided from a preset second time period, and the periods between each sliding window period are the same, and each sliding window period is pre-set with a corresponding sliding window weight.
[0187] As an example, the weight of the i-th sliding window period is a i and b i are pre-determined parameters.
[0188] Step f2, calculating the compactness of all sliding window periods based on the weights and key features corresponding to each sliding window period, the predetermined decay coefficient corresponding to the current iteration number, and the predetermined weights of each key feature.
[0189] Specifically, firstly, the normalized number of times of each sliding window period is calculated by the following formula:
[0190]
[0191] Among them, N i represents the normalized number of the i-th sliding window period, K j represents the jth key feature, M j represents the weight of the jth key feature, and s is the number of key features determined.
[0192] Then, the tightness of each sliding window period is calculated by the following formula:
[0193] f i =d l *log2(N i +1);
[0194]
[0195] Among them, f i represents the tightness of the i-th sliding window period, d l Indicates the number of decays corresponding to the ,th iteration number;
[0196] Then, the tightness of all sliding window periods is calculated by the following formula:
[0197]
[0198] Among them, f represents the compactness of all sliding window periods, and z is the total number of divided sliding window periods.
[0199] Step f3, determining the closeness between the seed number and the pairing number of the current iteration number based on the decay number corresponding to the current iteration number and the closeness of all sliding windows.
[0200] Specifically, the closeness f′ between the seed number and the pairing number of the current iteration can be obtained by calculating the product of the decay times corresponding to the current iteration and the closeness of all sliding windows of the current iteration, that is, f′=df.
[0201] Through the above solution, the members of the first initial family circle and the second initial family circle are corrected according to the closeness of the family relationship between the user numbers, which can further improve the accuracy of the family member identification result.
[0202] Based on the second embodiment of this application, please refer to Fig. 9 In the fourth embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 2 can be referred to the above description, and will not be described in detail later. On this basis, after the above-mentioned step S80, the family circle identification method further includes steps S820 to S840:
[0203] Step S820: extract features from the communication data of the user numbers in the converged family circle according to preset statistical feature extraction rules to obtain multiple statistical features.
[0204] Among them, the preset statistical features can be set according to actual conditions. As an example, the preset statistical feature extraction rules may include: extracting at least one feature under multiple dimensions such as user characteristics, service characteristics, and Internet behavior characteristics of each user number.
[0205] As an example, the following Table 2 shows the field information of some statistical features:
[0206] Table 2
[0207] Field Tags type Field Chinese Description phone_number int User Number gender Varchar2 gender age Char age cart_month Char Cartoon viewing time is_xiaoxinka Varchar2 Is the card user? ...... ...... ...... is_order_cart float Subscribe to cartoons is_student_cart float Campus Card
[0208] Step S830: Perform a weighted sum operation on multiple statistical features to obtain a score.
[0209] In the solution of this embodiment, the preset weights of various statistical features may be obtained, and a weighted sum operation may be performed based on the weights of various statistical features and the values of various statistical features to obtain the score corresponding to each user number.
[0210] Step S840: Determine the role of the user corresponding to the user number in the family based on the interval to which the score belongs.
[0211] In this embodiment, roles may include decision makers, children, and the elderly. Specifically, when determining family roles, firstly, based on the business characteristics, the decision makers in each family circle are identified according to the preset family decision maker identification rules and marked on the corresponding user numbers, and then the intervals to which the scores of other user numbers except the user numbers marked with the decision maker are determined to obtain the role labels corresponding to each user number. For example, assuming that when the score is in the range of 0 to 70, the corresponding user number is marked with a child label, and when the score is in the range of 70 to 100, the corresponding user number is marked with an elderly label.
[0212] Through the above implementation scheme, by calculating the scores of the statistical features of each user number to identify the role of the corresponding user number in the family circle, the member structure relationship of the family circle can be identified.
[0213] Based on the first embodiment of the present application, in the fifth embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction, and will not be repeated later. On this basis, the communication data also includes the service information of the user number, and after the above step S10, it also includes:
[0214] Based on the service information of each user number, the service characteristics of each user number are determined; then, based on the service characteristics of each user number, multiple user numbers with potential family relationships are identified through preset business rules to obtain a third initial family circle.
[0215] In this embodiment, the preset business rules may include: a. Determining whether there is a potential family relationship among user numbers handled under the same package (the customer's ID user number is the ID card). The same package here refers to a primary and secondary card integrated package, and the same package corresponds to multiple user numbers; b. Determining the potential family relationship among multiple packages of the same identity user number; c. Determining whether there is a potential family relationship among multiple user numbers of the same family card service; d. Determining whether there is a potential family relationship among multiple user numbers bound to the same family broadband service.
[0216] After determining multiple user numbers with family relationships, the multiple user numbers with family relationships are merged to obtain a third initial family circle. The third initial family circle may include one or more.
[0217] Furthermore, after determining the third initial family circle, the decision maker in the family circle can be identified by using a preset decision maker identification rule. The preset decision maker identification rule may include: using the person who handles the family card package service as the decision maker in the family circle; using the person who handles the family broadband service as the decision maker in the family circle.
[0218] Furthermore, after the above step S30, the method may also include: clustering user numbers whose corresponding traffic data in the same gateway meet preset home category conditions to obtain a first initial family circle; and merging the first initial family circle and the third initial family circle to obtain a merged family circle.
[0219] Specifically, the implementation steps and principles for integrating the first initial family circle and the third initial family circle are consistent with the implementation steps and principles for integrating the first initial family circle and the second initial family circle in the above embodiment, and are not described in detail here.
[0220] Based on the second embodiment of the present application, in the sixth embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above description, and will not be described in detail later. On this basis, the communication data also includes service information of each user number, and the family identification method further includes the following after the above step S10:
[0221] Based on the service information of each user number, the service characteristics of each user number are determined; based on the service characteristics of each user number, multiple user numbers with potential family relationships are identified through preset business rules to obtain a third initial family circle.
[0222] Correspondingly, after the above step S80, the method may further include: fusing the first initial family circle, the second initial family circle and the third initial family circle to obtain a merged family circle.
[0223] Specifically, the implementation steps and principles for integrating the first initial family circle, the second initial family circle, and the third initial family circle are consistent with the implementation steps and principles for integrating the first initial family circle and the second initial family circle in the above embodiment, and are not described in detail here.
[0224] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the family member identification method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.
[0225] This application also provides a family member identification device, please refer to Fig.10 , the family member identification device comprises:
[0226] The data acquisition module 10 is used to acquire the collected communication data; the communication data includes the flow data generated by each user number when passing through the gateway;
[0227] The classification module 20 is used to analyze the traffic data, obtain the online behavior classification result, and determine whether the traffic data meets the preset home category condition based on the online behavior classification result;
[0228] The determination module 30 is used to determine the users corresponding to the user numbers whose corresponding traffic data in the same gateway meet the preset home category condition as members in the same family circle.
[0229] The family member identification device provided by the present application adopts the family member identification method in the above embodiment, and can solve the technical problems of the family member identification method. Compared with the prior art, the beneficial effects of the family member identification device provided by the present application are the same as the beneficial effects of the family member identification method provided by the above embodiment, and other technical features in the family member identification device are the same as the features disclosed in the above embodiment method, which will not be repeated here.
[0230] The present application provides a family member identification device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the family member identification method in the above-mentioned embodiment one.
[0231] Reference below Fig.11 , which shows a schematic diagram of the structure of a family member identification device suitable for implementing the embodiment of the present application. The family member identification device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The family member identification device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0232] like Fig.11 As shown, the family member identification device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM) 1004. In RAM 1004, various programs and data required for the operation of the family member identification device are also stored. The processing device 1001, ROM 1002, and RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the family member identification device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a family member identification device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.
[0233] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.
[0234] The family member identification device provided by the present application adopts the family member identification method in the above embodiment, which can solve the technical problem of family member identification. Compared with the prior art, the beneficial effects of the family member identification device provided by the present application are the same as the beneficial effects of the family member identification method provided by the above embodiment, and other technical features in the family member identification device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.
[0235] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.
[0236] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0237] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, wherein the computer-readable program instructions are used to execute the family member identification method in the above-mentioned embodiment.
[0238] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.
[0239] The computer-readable storage medium may be included in the family member identification device; or may exist independently without being assembled into the family member identification device.
[0240] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the family member identification device, the family member identification device: obtains the collected communication data of the user number; the communication data includes the traffic data generated by each user number when passing through each gateway; analyzes the traffic data to obtain the Internet behavior classification result, and determines whether the traffic data meets the preset home category conditions based on the Internet behavior classification result; determines the users corresponding to the user numbers whose corresponding traffic data in the same gateway meets the preset home category conditions as members of the same family circle.
[0241] Computer program code for performing the operations of the present application may be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0242] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0243] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.
[0244] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned family member identification method, and can solve the technical problems of the family member identification method. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the family member identification method provided in the above-mentioned embodiment, and will not be repeated here.
[0245] The present application also provides a computer program product, including a computer program, which implements the steps of the family member identification method as described above when executed by a processor.
[0246] The computer program product provided in this application can solve the technical problem of family member identification. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the family member identification method provided in the above embodiment, which will not be repeated here.
[0247] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.
Claims
1. A method for identifying family members, characterized in that: The method includes: Acquire the collected communication data of the user number; the communication data includes the flow data generated by the user number when passing through each gateway; The traffic data is analyzed to obtain an Internet behavior classification result, and based on the Internet behavior classification result, it is determined whether the traffic data meets a preset home category condition; and it is determined that the users corresponding to the user numbers whose corresponding traffic data in the same gateway meets the preset home category condition are members of the same family circle.
2. The method according to claim 1, characterized in that The online behavior classification result includes the proportion of different types of online behaviors. The steps of analyzing the traffic data to obtain the online behavior classification result, and determining whether the traffic data meets the preset home category condition based on the online behavior classification result include: Inputting the traffic data into a pre-trained traffic classification model to obtain the proportion of different types of online behaviors; Based on the proportions of the different categories of online behaviors, it is determined whether the traffic data meets a preset home category condition.
3. The method according to claim 2, characterized in that The step of determining whether the traffic data meets the preset home category condition based on the proportion of the different categories of online behaviors includes: Determining an online behavior score based on the proportions of the different types of online behaviors and predetermined weights of the different types of online behaviors; Based on the comparison result of the online behavior score and the preset threshold, it is determined whether the traffic data meets the preset home category condition.
4. The method according to any one of claims 1 to 3, characterized in that The communication data also includes user information, signaling data and location information corresponding to the user number. After the step of acquiring the collected communication data of the user number, the method further includes: Constructing a first multi-dimensional feature of each of the user numbers based on the user information, signaling data, and location information corresponding to each of the user numbers; A plurality of user numbers having potential family relationships are identified through the first multi-dimensional features of each of the user numbers.
5. The method according to claim 4, characterized in that The location information includes a permanent location data set, and before the step of identifying a plurality of user numbers having a potential family relationship through the first multi-dimensional features of each of the user numbers, the step further includes: Calculating the similarity of the permanent location data sets between the user numbers to obtain the permanent location similarity between the user numbers; The step of identifying a plurality of user numbers having a potential family relationship by using the first multi-dimensional features of each of the user numbers comprises: Based on the first multi-dimensional features of each of the user numbers and the similarity of the permanent locations between the user numbers, construct a second multi-dimensional feature of the number pair; the number pair is two different user numbers; The second multi-dimensional features of the number pairs are input into a pre-trained social circle recognition model to determine whether there is a potential family relationship between the number pairs.
6. The method according to claim 5, characterized in that The step of calculating the similarity of the permanent location data sets between the user numbers to obtain the similarity of the permanent locations between the user numbers comprises: Mapping the resident location data set of each user number to a preset inner product space based on a preset kernel function to obtain the data point to which each user number is mapped; In the inner product space, the inner products between the data points to which the user numbers are mapped are calculated to obtain the resident position similarities between the user numbers.
7. The method according to claim 4, characterized in that After the step of identifying a plurality of user numbers having a potential family relationship through the first multi-dimensional features of each of the user numbers, the method further comprises: Clustering user numbers whose corresponding traffic data in the same gateway meet preset home category conditions to obtain a first initial home circle; Clustering multiple user numbers with potential family relationships to obtain a second initial family circle; The first initial family circle and the second initial family circle are merged to obtain a merged family circle.
8. The method according to claim 7, characterized in that The step of integrating the first initial family circle and the second initial family circle to obtain an integrated family circle includes: Merging the user numbers in the first initial family circle and the second initial family circle to obtain a candidate family circle; A seed number and a pairing number corresponding to the seed number are selected from the candidate family circle, and the family closeness between the seed number and the pairing number is determined; based on the family closeness, the corresponding pairing number is added to the family circle of the seed number to obtain the integrated family circle; wherein the seed number and the pairing number are both user numbers in the integrated family circle.
9. The method according to claim 8, characterized in that The step of selecting a seed number and a pairing number corresponding to the seed number from the candidate family circle, determining the family closeness between the seed number and the pairing number, and adding the corresponding pairing number to the family circle of the seed number based on the family closeness to obtain the integrated family circle includes: Selecting a seed number corresponding to an initial number of iterations from the candidate family circle, and initializing the family circle to be expanded of the seed number; The family circle to be expanded of the seed number is subjected to multiple iterative processing, and each iterative processing includes: selecting each paired number corresponding to the seed number in the candidate family circle, and processing the communication data of the seed number and the communication data of each paired number according to a preset key feature extraction rule to obtain the key features between the seed number and each paired number; determining the family closeness between the seed number and each paired number based on the key features between the seed number and each paired number; adding the paired number whose family closeness is greater than a preset closeness threshold to the family circle to be expanded of the seed number, and selecting the paired number added to the family circle to be expanded as the seed number corresponding to the next number of iterations; When it is detected that the number of number members in the family circle to be expanded of the seed number is greater than or equal to a preset limit threshold, the iterative processing is stopped, and the family circle to be expanded of the seed number when the iterative processing is stopped is determined, and the merged family circle is determined based on the family circle to be expanded of the seed number when the iterative processing is stopped.
10. The method according to claim 7, characterized in that After the step of fusing the first initial family circle and the second initial family circle to obtain a merged family circle, the method further comprises: Extracting features from the communication data of the user number in the integrated family circle according to a preset statistical feature extraction rule to obtain a plurality of statistical features; Performing a weighted sum operation on the multiple statistical features to obtain a score; Based on the interval to which the score belongs, the role of the user corresponding to the user number in the family is determined.
11. The method according to claim 1, characterized in that The communication data also includes service information of the user number. After the step of acquiring the collected communication data of the user number, the method further includes: Determining service characteristics of each of the user numbers based on the service information of each of the user numbers; Based on the service characteristics of each of the user numbers, a plurality of user numbers with potential family relationships are identified through preset service rules to obtain a third initial family circle; After the step of determining that the user corresponding to the user number whose corresponding traffic data in the same gateway meets the preset home category condition is a member of the same family circle, the method further includes: Clustering user numbers whose corresponding traffic data in the same gateway meet preset home category conditions to obtain a first initial home circle; The first initial family circle and the third initial family circle are merged to obtain a merged family circle.
12. The method according to claim 11, characterized in that The step of integrating the first initial family circle and the third initial family circle to obtain an integrated family circle includes: Merging the user numbers in the first initial family circle and the third initial family circle to obtain a candidate family circle; A seed number and a pairing number corresponding to the seed number are selected from the candidate family circle, and the family closeness between the seed number and the pairing number is determined; based on the family closeness, the corresponding pairing number is added to the family circle of the seed number to obtain the integrated family circle; wherein the seed number and the pairing number are both user numbers in the integrated family circle.
13. A family member identification device, characterized in that: include: A data acquisition module, used to acquire the collected communication data of the user number; the communication data includes the flow data generated by the user number when passing through each gateway; A classification module, used to analyze the traffic data to obtain an Internet behavior classification result, and determine whether the traffic data meets a preset home category condition based on the Internet behavior classification result; The determination module is used to determine that the users corresponding to the user numbers whose corresponding traffic data in the same gateway meet the preset home category conditions are members of the same family circle.
14. A family member identification device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the family member identification method according to any one of claims 1 to 12.
15. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the family member identification method according to any one of claims 1 to 12 are implemented.
16. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the family member identification method according to any one of claims 1 to 12 are implemented.
