Family member identification method, apparatus, medium, and electronic device

Abstract

The application relates to the fields of big data and artificial intelligence, and discloses a family member identification method and device, a medium and an electronic device. The method comprises the following steps: acquiring data associated with a mobile phone number from multiple data sources respectively; performing family circle identification based on a family circle identification method corresponding to each data source according to the data acquired from the data source, to obtain a type of family circle corresponding to the data source; establishing a family graph network according to the types of family circles; iteratively performing an adjusting family graph network step, and when the information entropy of the family graph network no longer changes after the adjusting family graph network step is performed this time, taking the adjusted family graph network as a final family graph network, and taking nodes belonging to the same family circle in the final family graph network as family members of a family. The method can more accurately divide the family circle, and further improves the accuracy of family member identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data and artificial intelligence, and particularly relates to a family member identification method and device, a medium and an electronic device. BACKGROUND

[0002] The family market is one of the key competitive markets in the communication industry. With the development of 5G, full-service and integrated packages, the family market is becoming more and more important, and is the main battlefield for business development and precise marketing, which has a very important significance for customer management. However, there are few solutions to identify family members at present. Even if there are corresponding solutions, they often have the defects of single identification method and low identification accuracy. SUMMARY

[0003] In the technical field of big data and artificial intelligence, in order to solve the above technical problems, the purpose of the present application is to provide a family member identification method and device, a medium and an electronic device.

[0004] According to an aspect of the present application, a family member identification method is provided, which comprises:

[0005] Respectively acquiring data associated with a mobile phone number from a plurality of data sources;

[0006] According to the data acquired from each data source, performing family circle identification based on the family circle identification method corresponding to the data source to obtain a type of family circle corresponding to the data source, the family circle comprising at least one mobile phone number;

[0007] According to various types of family circles, establishing a family graph network, the family graph network comprising nodes and edges connecting two nodes, at least one node in the family graph network being a mobile phone number, the edges being used to indicate that the two nodes connected by the edges belong to one family circle, and the weight of the edge being associated with the family circle identification method for identifying that the two nodes connected by the edge belong to the same family circle;

[0008] Iteratively performing an adjusting family graph network step, according to the information entropy of the family graph network no longer changing after performing the adjusting family graph network step this time, taking the adjusted family graph network as a final family graph network, and taking the nodes belonging to the same family circle in the final family graph network as the family members of one family, the adjusting family graph network step comprising:

[0009] The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative.

[0010] Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

[0011] According to another aspect of this application, a family member identification device is provided, the device comprising:

[0012] The acquisition module is configured to retrieve data associated with mobile phone numbers from multiple data sources.

[0013] The identification module is configured to identify a family circle based on the family circle identification method corresponding to each data source, according to the data obtained from each data source, to obtain a family circle corresponding to the data source, wherein the family circle includes at least one mobile phone number;

[0014] The module is configured to build a family graph network based on various family circles. The family graph network includes nodes and edges connecting two nodes. At least one node in the family graph network is a mobile phone number. The edges are used to indicate that the two connected nodes belong to the same family circle. The weight of the edge is associated with a family circle identification method that identifies that the two nodes connected by the edge belong to the same family circle.

[0015] The adjustment module is configured to iteratively execute the family graph network adjustment steps. After completing this adjustment step, the information entropy of the family graph network no longer changes. The adjusted family graph network is then taken as the final family graph network, and nodes belonging to the same family circle in the final family graph network are considered as family members of one family. The family graph network adjustment steps include:

[0016] The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative.

[0017] Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

[0018] According to another aspect of this application, a computer-readable program medium is provided that stores computer program instructions, which, when executed by a computer, cause the computer to perform the method described above.

[0019] According to another aspect of this application, an electronic device is provided, the electronic device comprising:

[0020] processor;

[0021] A memory storing computer-readable instructions that, when executed by the processor, implement the method described above.

[0022] The technical solutions provided by the embodiments of this application may include the following beneficial effects:

[0023] The family member identification method provided in this application first obtains data from multiple data sources, then establishes a family circle corresponding to each data source, and builds a family graph network based on each family circle. Family members are determined based on the family graph network, which is equivalent to constructing a hybrid multimodal integrated identification model for family members. By integrating data from multiple data sources and multiple family circle identification methods, a comprehensive family relationship is formed, avoiding the loss of family members and ensuring the accuracy of family member identification. On this basis, by iteratively executing the step of adjusting the family graph network, the identified family relationships are further eliminated, achieving a more precise division of family circles and further improving the accuracy of family member identification.

[0024] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0026] Figure 1 This is a schematic diagram of a system architecture for a family member identification method according to an exemplary embodiment;

[0027] Figure 2 This is a flowchart illustrating a family member identification method according to an exemplary embodiment;

[0028] Figure 3 This is illustrated according to an exemplary embodiment. Figure 2 A flowchart detailing step 220 is provided.

[0029] Figure 4 This is a schematic diagram illustrating the establishment of a family circle based on grid spatial data according to an exemplary embodiment;

[0030] Figure 5 This is a schematic diagram illustrating indicators selected according to feature type according to an exemplary embodiment;

[0031] Figure 6 This is a schematic diagram illustrating the identification of a user's nighttime residence in a residential cell, according to an exemplary embodiment.

[0032] Figure 7 This is a schematic diagram illustrating the process of establishing and adjusting a family graph network according to an exemplary embodiment;

[0033] Figure 8 This is a schematic diagram of a family graph network according to an exemplary embodiment;

[0034] Figure 9 This is a schematic diagram illustrating the results of family member identification and the family member relationships removed by the family member identification method according to an exemplary embodiment;

[0035] Figure 10 This is a block diagram illustrating a family member identification device according to an exemplary embodiment;

[0036] Figure 11 This is an example block diagram of an electronic device implementing the above-described family member identification method, according to an exemplary embodiment;

[0037] Figure 12 This is a program product that implements the above-described family member identification method, as shown in an exemplary embodiment. Detailed Implementation

[0038] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0039] Furthermore, the accompanying drawings are merely illustrative of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0040] This application first provides a method for identifying family members. The family is a basic unit of social life, and family member identification is a method for determining members who belong to the same family. In the family member identification method provided in this application, family members are represented by mobile phone numbers. The family member identification method provided through the embodiments of this application can achieve accurate and comprehensive identification of family members.

[0041] The implementation terminal of this application can be any device with computing capabilities. This device can be connected to external devices to receive or send data. Specifically, it can be a portable mobile device, such as a smartphone, tablet, laptop, PDA (Personal Digital Assistant), etc., or a fixed device, such as a computer, field terminal, desktop computer, server, workstation, etc., or a collection of multiple devices, such as the physical infrastructure of cloud computing or a server cluster.

[0042] Optionally, the implementation terminal of this application can be a server or a cloud computing physical infrastructure.

[0043] Figure 1 This is a schematic diagram of a system architecture for a family member identification method according to an exemplary embodiment. Figure 1As shown, the system architecture includes a data source, a model layer, and results. The data source includes telecommunications service data, DPI (Deep Packet Inspection) internet access data, third-party data, and call interaction data. The model layer includes rule-based statistical algorithms, clustering algorithms + rule statistics, rule-based statistical algorithms, and DNN deep learning. These methods obtain preliminary family relationships for users. The model layer also includes 5G SA signaling data and fixed-line Wi-Fi data, which are used to identify users' nighttime frequented cells. Based on the preliminary family relationships and users' nighttime frequented cells, the model layer also derives a graph computation model, and the identification result is the final family circle. The final family circle can include family IDs, family members from the same network, and family members from different networks. Here, "different networks" refers to networks of different entities, such as networks of different operators. For ease of understanding, the identification results are usually presented in a display module, allowing the results to be displayed and output to the user.

[0044] Figure 2 This is a flowchart illustrating a family member identification method according to an exemplary embodiment. The family member identification method provided in this embodiment can be executed by a server, such as... Figure 2 As shown, it includes the following steps:

[0045] Step 210: Obtain data associated with the mobile phone number from multiple data sources.

[0046] Multiple data sources can be categorized as telecommunications service data, DPI (Data Points Per Intake) internet access data, third-party data, and call interaction data. Telecommunications service data can include data from the telecommunications business support system (B / O) and service platform data. The B / O domain is the data domain for the business support system, and the O / O domain is the data domain for the operation support system. The B / O domain contains user data and service data, such as user consumption habits, terminal information, ARPU (Average Revenue Per User) groupings, service content, and target audience. The O / O domain contains network data, such as signaling, alarms, faults, and network resources.

[0047] As mentioned earlier, family members are identified by their mobile phone numbers; therefore, it is necessary to obtain data associated with these mobile phone numbers to identify family members. This data can be directly or indirectly linked to the mobile phone numbers.

[0048] In this embodiment of the application, the telecommunications BO domain data may include package information, call logs, CRM (Customer Relationship Management) information, ID numbers, customer contacts, order contacts, etc. The business platform may include data related to YiPay, complaint and consultation, and ITV platform. This data is related to the specific business of the telecommunications company.

[0049] DPI (Deep Packet Inspection) is a data packet-based deep inspection technology that performs deep inspections on different network application layer payloads (such as HTTP and DNS).

[0050] Call interaction data is data related to calls, such as call time, the other party's mobile phone number, and other call record information.

[0051] Third-party data refers to data provided by companies or entities other than telecom operators. For example, it could be personal data collected by a neighborhood committee; or data from data companies or the main companies behind various applications.

[0052] Step 220: Based on the data obtained from each data source, perform family circle identification based on the family circle identification method corresponding to the data source to obtain a type of family circle corresponding to the data source.

[0053] The family circle includes at least one mobile phone number.

[0054] Each family circle can include multiple family circles. Different data sources can use the same or different family circle identification methods for identification. The final established family circle consists of the family members within a family identified based on the corresponding family circle identification method.

[0055] Figure 3 This is illustrated according to an exemplary embodiment. Figure 2 A flowchart detailing step 220. (See attached flowchart.) Figure 3 As shown, step 220 may specifically include the following steps:

[0056] Step 221: Identify the telecommunications service-related data according to the predetermined business rules to obtain the first type of family circle.

[0057] The rules for pre-ordering services may include the following: a) Grouping mobile numbers under the same plan as a family group (the customer's ID number is their national ID card). Here, "same plan" refers to a primary and secondary SIM card fusion plan, where the same plan corresponds to multiple mobile numbers; b) Grouping multiple plans under the same ID number as a family group; c) Grouping the customer's contact person and the corresponding customer as a family group; d) Grouping the order contact person and the service number used for the service as a family group; e) Grouping the complaint number and the number of the service being complained about as a family group. Here, the complaint number is the mobile number used to call customer service, and the number of the service being complained about is the mobile number used for warranty service; f) Grouping the YiPay registration number and the customer with the same ID number used for YiPay registration as a family group; g) Excluding: family relationships with no network usage records for three months. Network usage records refer to call records between different operators.

[0058] Step 222: Obtain the corresponding feature data from the depth packet detection data of each broadband.

[0059] The feature data includes feature values ​​corresponding to pre-constructed features.

[0060] In most cases, broadband is also one of the important services of operators, and broadband is linked to mobile phone numbers. Therefore, family members can be identified based on broadband usage. As mentioned earlier, deep packet inspection data is DPI (Deep Packet Inspection) internet access data. For DPI internet access data, this application embodiment uses clustering algorithms and rule statistical algorithms to identify family members.

[0061] First, through in-depth analysis of DPI internet access data, the terminals and phone numbers used for broadband internet access are identified. Specifically, the UA and COOKIES information in the DPI internet access data is analyzed in depth to obtain the terminals and phone numbers used for broadband internet access.

[0062] We analyze the number of terminals connected to the same broadband network over a week and the number of broadband connections a single terminal makes over a week to construct features. Specifically, we construct features such as the number of terminals connected to the broadband network over a week, the number of connections made by each terminal, the connection duration, the number of connections made during the day, the number of connections made at night, the number of terminal connections made during the day, and the number of terminal connections made at night. Based on these features, we can obtain corresponding characteristic data.

[0063] In one embodiment, before obtaining the corresponding feature data from the deep packet inspection data of each broadband, the method further includes: determining, based on the deep packet inspection data, that the number of broadband connections connected to the target terminal within a specified time period is greater than a specified number, and removing the data corresponding to the target terminal from the deep packet inspection data.

[0064] Specifically, the specified time period can be one week, and the specified number can be 5. If the number of broadband connections connected to the target terminal within the specified time period is greater than 5, it indicates that the user of the target terminal moves frequently, and its data is not suitable for use as data to identify home broadband.

[0065] In this embodiment of the application, the accuracy of identifying home broadband is improved by removing the data corresponding to the target terminal.

[0066] In one embodiment, after obtaining the corresponding feature data from the deep packet inspection data of each broadband, the method further includes: based on the deep packet inspection data, removing mobile phone numbers whose number of broadband connections within a predetermined time period exceeds a predetermined number.

[0067] The booking period can be one week, and the number of bookings can be 5.

[0068] If a mobile phone number connects to too many broadband connections within a certain period, it indicates that the user of that number frequently moves around. Even if the number is identified as belonging to a family member, the identification may not be accurate. In this embodiment, by eliminating such mobile phone numbers, the accuracy of family member identification is ensured.

[0069] Step 223: Perform clustering processing on the feature data of each broadband, and determine the home broadband from each broadband based on the clustering results.

[0070] Feature data can be clustered using various algorithms such as K-means and DBSCN.

[0071] Specifically, clustering can distinguish the types of broadband, which can include home broadband and work broadband.

[0072] Step 224: Based on the deep packet inspection data, determine the mobile phone numbers connected to the same home broadband network as the second type of home network.

[0073] If multiple mobile phone numbers are connected to the same home broadband, it means that these mobile phone numbers belong to a family member.

[0074] Step 225: Based on the grid space data, the mobile phone numbers corresponding to the same grid space address are designated as the third type of family circle.

[0075] Each of the aforementioned family clusters consists of mobile phone numbers. The grid spatial data is third-party data, which may be data collected by grid workers from the street office, including each address and its corresponding mobile phone number.

[0076] Based on grid spatial data, family members are also identified by using rule-based statistical algorithms.

[0077] Figure 4 This is a schematic diagram illustrating the creation of a family circle based on grid spatial data according to an exemplary embodiment. Figure 4 As shown, based on grid spatial data, family circles can be established by identifying users from other networks and users from the same network. A family circle can include only users from other networks, users from both other networks and users from the same network, or only users from the same network. Specifically, a grid cell corresponds to a grid spatial address, which in practice may be a room. All users under the same grid cell are considered a family. Based on the established family relationship table, if a certain number from another network already exists, families are associated and merged based on that number, for example... Figure 4 The left side shows 150****3421 and 185****2628, which are users on different networks within the same room. If these two users are associated with other users in the existing family relationship table, they can be merged. Family members can also be identified based on users within the same network. For users on different networks within the same room, relationships are established through their network numbers, merging them into one family. Specifically, if CRM data determines that the broadband in a room is KDYX55****, and the mobile phone number bound to this broadband is 153****4597, and if grid space data determines that the room also includes a mobile phone number on a different network, 150****3421, then… Figure 4 The embodiment can consider 153****4597 and 150****3421 as a family, that is, according to Figure 4 In this example, the mobile phone numbers of users on different networks in the same room and the mobile phone numbers under the broadband service packages in the same room are considered as belonging to the same household.

[0078] In other embodiments of this application, family circles can also be established based on data from courier companies and e-commerce platforms.

[0079] In one embodiment, after identifying mobile phone numbers corresponding to the same grid space address as a third type of family circle based on grid space data, the step of obtaining a family circle corresponding to the data source based on the family circle identification method corresponding to the data source, according to data obtained from each data source, further includes: extracting family relationship data and call feature data corresponding to mobile phone numbers in the first, second, and third types of family circles, wherein the call feature data includes feature values ​​corresponding to multiple call features respectively; training a deep neural network model based on the call feature data to obtain a family member identification model, and performing family member identification based on the family member identification model to obtain mobile phone number pairs; and adjusting the first, second, and third types of family circles according to the mobile phone number pairs.

[0080]

[0081] Table 1

[0082] Please refer to Table 1. After testing and evaluation, it was found that the recall rate of the DNN deep learning method was 78% and the coverage rate was 69%, which is significantly better than other traditional algorithms.

[0083] Deep Neural Networks (DNN) models are models built based on the connections between neurons. In other embodiments of this application, other machine learning or deep learning models may also be used.

[0084] Specifically, indicators can be selected from dimensions such as stability, continuity, frequency, volatility, clustering of call time periods, mutual contacts, and call duration to construct call characteristics. Figure 5 This is a schematic diagram illustrating indicators selected based on feature types according to an exemplary embodiment. Feature type is equivalent to a dimension for indicator selection, and the indicator is the selected call feature. Among them, stability refers to the stability of the call relationship, where members tend to maintain contact, for example, there are calls most often within a day or month; an example indicator is the percentage of days with calls within a month. Persistence refers to the longevity of the call relationship; examples of indicators are the longest consecutive call days, the longest consecutive call weeks, and the longest consecutive call months. Frequency refers to frequent call behavior among family members; examples of indicators are the average number of calls per day and the ranking of monthly call frequency. Volatility refers to the higher volatility of call activity on weekdays compared to rest days / significantly higher volatility on holidays compared to weekdays; an example indicator is the standard deviation of rest days and holidays / the standard deviation of working hours on weekdays. Call time clustering refers to the tendency for calls between family members to occur outside of working hours (on the way to work in the morning, during lunch break, and on the way home from work) on weekdays; an example indicator is the percentage of calls made outside of working hours on weekdays. Shared contacts refer to analyzing the difference between family relationships and other relationships from the perspective of the number of shared contacts.

[0085] In this embodiment of the application, a family member identification model based on a deep neural network model is constructed. The data used to train the model is based on data corresponding to various family circles. The generated family circles are then adjusted based on the mobile phone number pairs identified by the family member identification model, which further improves the accuracy of family member identification.

[0086] In other embodiments of this application, a fourth type of family circle can also be constructed separately using the identification results of the family member identification model.

[0087] In one embodiment, before extracting call feature data of mobile phone numbers in the first, second, and third family circles, the method further includes: determining the nighttime permanent cell corresponding to each mobile phone number based on the signaling data and broadband installation address of the mobile phone number; the extraction of family relationship data and call feature data corresponding to the mobile phone numbers in the first, second, and third family circles includes: extracting family relationship data of mobile phone numbers in the first, second, and third family circles, and adjusting the family relationship data based on the inconsistency of the nighttime permanent cell corresponding to the mobile phone numbers within the same family circle; and extracting call feature data corresponding to each mobile phone number.

[0088] Figure 6 This is a schematic diagram illustrating the identification of a user's nighttime residence cell according to an exemplary embodiment. For details, please refer to... Figure 6 For users within the same network, the system identifies their nighttime residential communities based on signaling data and broadband installation addresses. This signaling data can be, for example, 4G or 5G signaling data. For users from other networks, the system identifies their nighttime residential communities based on their Wi-Fi location and network affiliation model. Finally, when outputting the family circle data, an attribute indicating whether users with family relationships reside together at night can be added to all identified users. Users residing in the same community have this attribute value of 1; otherwise, the value is 0.

[0089] In this embodiment, by adjusting the family relationship data based on the nighttime residential area of ​​the mobile phone number, non-family members such as relatives, colleagues, and friends can be filtered out, ensuring the accuracy of the data and thus improving the accuracy of the trained family member identification model.

[0090] In one embodiment, determining mobile phone numbers connected to the same home broadband network as a second type of home network based on the deep packet inspection data includes: determining that the mobile phone numbers connected to the home broadband network are from different networks based on the deep packet inspection data, and using the third-party application identifier in the deep packet inspection data and the mobile phone numbers connected to the home broadband network as the second type of home network; the method further includes: identifying the nighttime resident cell based on the Wi-Fi location corresponding to the third-party application identifier in the deep packet inspection data.

[0091] Specifically, for cross-network numbers that cannot be identified, WeChat OPENID can be used as a substitute to identify the cross-network user. Here, WeChat OPENID is the identifier of a third-party application. WeChat OPENID is parsed from the cookie information in the DPI data. Therefore, this embodiment can also effectively identify family members of cross-network users and accurately identify the communities where cross-network users frequently reside at night.

[0092] Step 230: Establish a family graph network based on various family circles.

[0093] The family graph network includes nodes and edges connecting two nodes. At least one node in the family graph network is a mobile phone number. The edges are used to indicate that the two connected nodes belong to the same family circle. The weight of the edge is associated with a family circle identification method that identifies the two nodes connected by the edge as belonging to the same family circle.

[0094] The weight of an edge is its attribute value. Since each family circle identification method corresponds to a data source, the weight of the edge also corresponds to the data source.

[0095] In the established family graph network, each node corresponds to a user or family member. The node may be a mobile phone number or the third-party application identifier mentioned earlier.

[0096] If four family circles are established based on four types of data sources, then the analytic hierarchy process (AHP) can be used to assign weights w1, w2, w3, and w4 to the family relationships in each family circle. Therefore, the edge weights can be: w = w1 + w2 + w3 + w4.

[0097] In one embodiment, the method further includes: adjusting the weights of edges between nodes in the family graph network based on the inconsistency of the nighttime residential communities corresponding to mobile phone numbers within the same family circle.

[0098] Specifically, the final edge weight can be: w = (w1 + w2 + w3 + w4) * w5. Here, w5 is an adjustment coefficient. If users live in the same community at night, w5 = 1; otherwise, w5 is a minimum value, such as 0 or 0.01. Setting w5 to 0.01 preserves traces of family relationships in the family graph network. This achieves the purpose of adjusting the edge weight.

[0099] In this embodiment, by adjusting the edge weights according to the user's nighttime residence, the family graph network can more accurately represent family relationships.

[0100] Figure 8 This is a schematic diagram of a family graph network according to an exemplary embodiment.Figure 8 As shown, the family graph network is equivalent to a graph model, which includes several nodes such as A, B, and C, and the numbers on the edges between the nodes are the weights.

[0101] In one embodiment, the method further includes: deleting edges between corresponding nodes in the family graph network based on the inconsistency of the nighttime residential communities corresponding to mobile phone numbers within the same family circle.

[0102] Since the neighborhoods corresponding to mobile phone numbers may differ at night, the nodes associated with these numbers are often not related as family members. In this embodiment, the accuracy of the established family graph network is ensured by deleting the edges between these nodes.

[0103] Step 240: Iteratively execute the family graph network adjustment step. After completing this family graph network adjustment step, the information entropy of the family graph network no longer changes. The adjusted family graph network is taken as the final family graph network, and nodes belonging to the same family circle in the final family graph network are taken as family members of a family.

[0104] The step of adjusting the family graph network includes:

[0105] The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative.

[0106] Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

[0107] When the maximum information gain is negative, if the node has already been added to the family circle, it needs to be returned to its original family; if the node has not been added to the family circle, no additional operation is required.

[0108] Specifically, after performing the adjustment step for the family graph network, if the information entropy of the family graph network is the same as that before performing the adjustment step, then it can be considered that the information entropy of the family graph network no longer changes after performing the adjustment step.

[0109] Figure 7 This is a schematic diagram illustrating the process of establishing and adjusting a family graph network according to an exemplary embodiment.

[0110] like Figure 7 As shown, the process includes the following steps:

[0111] 1. Construct a family graph network based on the obtained phone number pairs with family relationships.

[0112] Using the identified mobile phone numbers and third-party application identifiers within the same family circle as nodes, edges are established between nodes based on the family relationships corresponding to the nodes. The weights between each edge can be set as w = (w1 + w2 + w3 + w4) * w5.

[0113] 2. Calculate the information gain value of all nodes connected to it.

[0114] Initially, each node is treated as a family circle. Each node is then sequentially added to each of its neighboring family circles, and the information gain value is calculated using the following formula:

[0115]

[0116] in, K represents the information gain value. i,j K is the sum of the weights of the edges from node i to the j-th family circle. i Let ∑ be the sum of the weights of all edges pointing to node i. tot The sum of the weights of the edges connected to nodes within the family circle is given by , where m is the sum of the weights of all edges in the family graph network.

[0117] 3. Nodes with the maximum information gain value and a gain value greater than 0 are merged.

[0118] After adding a node to each of its neighboring family circles, if adding the node to a family circle produces the maximum information gain value, and this information gain value is greater than 0, then the node is merged with that family circle.

[0119] 4. Traverse all nodes in the family graph network to form a new graph.

[0120] The above operation is repeated for each node until the family relationships of all nodes no longer change, resulting in a new family graph network. As the nodes are traversed, a family circle may contain multiple nodes. A family circle can be considered as a supernode, and the edges between nodes in the family circle correspond to a self-loop edge of the supernode.

[0121] 5. Compare whether the information entropy before and after the merger is equal.

[0122] After each round of steps 2-4, the information entropy is calculated. The information entropy determined before each round of steps 2-4 is the information entropy before node merging; the information entropy determined after each round of steps 2-4 is the information entropy after node merging. The two entropies are checked for equality. If they are equal, the process ends; otherwise, steps 2-4 continue.

[0123] Information entropy can be calculated using the following formula:

[0124]

[0125] Where Q is the information entropy, and m is the sum of the weights of all edges in the family graph network; A ij K represents the weight of the edge between node i and node j; i K is the sum of the weights of all edges pointing to node i; j It is the sum of the weights of all edges pointing to node j; δ(C i C j ) is used to determine whether node i and node j belong to the same family. When node i and node j are in the same family, δ(C) i C j The value of ) is 0; otherwise, δ(C) i C j The value of ) is 1.

[0126] After obtaining the final family graph network, the nodes of each family circle in the final family graph network can be output as family members of the identified family.

[0127] The inventors of this application also conducted sampling verification of the final identification results. For the finally identified family members, a sampling method was used for verification. An average sample was taken from groups such as government agencies, enterprises, and merchants, with a total of 1000 data samples taken from the local network. The verification showed that the accuracy rate for identifying family members reached 92%, and the coverage rate reached 81%, both significantly improved compared to traditional methods.

[0128] Figure 9 This is a schematic diagram illustrating the results of family member identification and the family member relationships removed by the family member identification method, according to an exemplary embodiment. Figure 9 In the table, the left side shows the results of family member identification, and the right side shows the family member relationships that were removed through the family member identification method.

[0129] After identification and verification, the following conclusions can be drawn:

[0130] 1. 153xxxx4551 and 153xxxx4035 are numbers under the same service plan, but their nighttime base station locations are different, leading to their exclusion by the algorithm. The investigation results confirm that 153xxxx4551 gave the number 153xxxx4035 to his nephew.

[0131] 2. Based on call characteristics, 187xxxx5021 and 135xxxx6795 strongly suggest a family relationship. The number 135xxxx6795 is from a different network, making it impossible to directly determine the user's nighttime location through signaling data. However, by examining the user's Wi-Fi broadband address, their nighttime location was determined to be in another city, indicating a kinship relationship, which aligns with the facts.

[0132] 3. User wx1235678xx12 is the younger sister of 153xxxx4551. It is also a different network number. They usually do not communicate by phone, but mainly through WeChat. Through in-depth analysis of DPI, we supplemented the family members with the WeChat OPENID instead of the mobile phone number.

[0133] In summary, the family member identification method provided in this application constructs a multimodal ensemble algorithm to comprehensively model telecommunications BO domain data, business platform data, DPI internet access data, call interaction circles, and third-party data, forming a comprehensive family relationship and avoiding the problem of lost family members. It integrates multiple algorithms and constructs a graph computation model based on information entropy, performing secondary filtering on the comprehensive family member data while ensuring the accuracy of family member identification. It employs various business-integrated methods for dividing positive samples, such as based on packages, contacts, and business platforms (e.g., YiPay, complaints and inquiries), to expand the positive sample while eliminating family members who do not live together, effectively preventing sample imbalance. Furthermore, it effectively labels cross-network data based on WeChat OPENID, expanding cross-network family members, and effectively determines cross-network location based on the cross-network affiliation model, improving the accuracy and recall rate of family member identification. Finally, it uses 5G signaling data to identify the nighttime residential areas of family members, and by setting a small edge weight threshold, it can further filter out interference factors such as relatives, friends, and colleagues.

[0134] The family member identification method provided in this application can also be applied to other business scenarios. For example, from the perspective of the family, it can comprehensively depict the composition and usage of family communication products, accurately identify the product space, value space, and terminal lifecycle of each family, and lay the foundation for forming a precise marketing strategy tailored to each household. It can further explore business opportunities, including user space (family integration or defection of family members from different networks), product space (what other products can be developed within the family), and value space (whether the telecommunications products within the family can be upgraded or transferred).

[0135] In the context of a competitive telecommunications market with full-service offerings, the solutions proposed in this application can facilitate a shift in telecommunications services from a "product-centric" to a "customer-centric" approach, enabling the implementation of customer-demand-oriented market operation strategies. By analyzing the characteristics of household users, segmentation of household users can be achieved, laying the foundation for refined marketing, saving resource costs, and improving marketing accuracy.

[0136] In addition to supporting traditional telecommunications services, home-based data is also highly valuable when collaborating with other industries, such as:

[0137] The real estate industry can leverage family-based data to build housing demand models by family unit, considering factors such as family income, family size, property information, presence of preschool-aged children, and presence of marriageable family members. This model can assist in precise marketing within the real estate industry. Other industries with potential for collaboration include tourism, automotive, finance, and education.

[0138] This application also provides a family member identification device, and the following are embodiments of the device of this application.

[0139] Figure 10 This is a block diagram illustrating a family member identification device according to an exemplary embodiment. Figure 10 As shown, the device 1000 includes:

[0140] The acquisition module 1010 is configured to acquire data associated with mobile phone numbers from multiple data sources.

[0141] The identification module 1020 is configured to identify a family circle based on the family circle identification method corresponding to each data source, according to the data obtained from each data source, to obtain a family circle corresponding to the data source, wherein the family circle includes at least one mobile phone number.

[0142] The module 1030 is configured to establish a family graph network based on various family circles. The family graph network includes nodes and edges connecting two nodes. At least one node in the family graph network is a mobile phone number. The edges are used to indicate that the two connected nodes belong to the same family circle. The weight of the edge is associated with a family circle identification method that identifies that the two nodes connected by the edge belong to the same family circle.

[0143] The adjustment module 1040 is configured to iteratively execute the family graph network adjustment step. After completing this adjustment step, the information entropy of the family graph network no longer changes. The adjusted family graph network is then taken as the final family graph network, and nodes belonging to the same family circle in the final family graph network are considered as family members of one family. The family graph network adjustment step includes:

[0144] The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative.

[0145] Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

[0146] According to a third aspect of this application, an electronic device capable of implementing the above-described method is also provided.

[0147] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0148] The electronic device 1100 according to this embodiment of the present application will now be described with reference to 11. Figure 11 The electronic device 1100 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0149] like Figure 11As shown, the electronic device 11100 is manifested in the form of a general-purpose computing device. The components of the electronic device 1100 may include, but are not limited to: at least one processing unit 1110, at least one storage unit 1120, and a bus 1130 connecting different system components (including storage unit 1120 and processing unit 1110).

[0150] The storage unit stores program code, which can be executed by the processing unit 1110, causing the processing unit 1110 to perform the steps described in the "Embodiment Method" section above according to various exemplary embodiments of this application.

[0151] Storage unit 1120 may include readable media in the form of volatile storage units, such as random access memory (RAM) 1121 and / or cache memory 1122, and may further include read-only memory (ROM) 1123.

[0152] Storage unit 1120 may also include a program / utility 1124 having a set (at least one) program module 1125, such program module 1125 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0153] Bus 1130 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0154] Electronic device 1100 can also communicate with one or more external devices 1300 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 1100, and / or with any device that enables electronic device 1100 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be made via input / output (I / O) interface 1150, such as communication with display unit 1140. Furthermore, electronic device 1100 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 1160. As shown, network adapter 1160 communicates with other modules of electronic device 1100 via bus 1130. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 1100, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0155] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the method according to the embodiments of this application.

[0156] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible embodiments, various aspects of this application may also be implemented as a program product comprising program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of this application described in the "Exemplary Methods" section above.

[0157] refer to Figure 12 As shown, a program product 1200 for implementing the above-described method according to an embodiment of this application is described. It may employ a portable compact disc read-only memory (CD-ROM) and include program code, and can run on a terminal device, such as a personal computer. However, the program product of this application is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0158] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0159] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0160] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0161] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0162] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of this application, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0163] It should be understood that this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for identifying family members, characterized in that, The method includes: Data associated with mobile phone numbers is retrieved from multiple data sources. Based on the predetermined business rules, the relevant data of telecommunications services are identified to obtain the first type of family circle; Obtain corresponding feature data from the deep packet inspection data of each bandwidth, wherein the feature data includes feature values ​​corresponding to pre-constructed features; Cluster the feature data of each broadband, and determine the home broadband from each broadband based on the clustering results; Based on the deep packet inspection data, mobile phone numbers connected to the same home broadband are identified as the second type of home network; Based on grid space data, mobile phone numbers corresponding to the same grid space address are designated as the third type of family circle, and each family circle includes at least one mobile phone number; Based on various family circles, a family graph network is established. The family graph network includes nodes and edges connecting two nodes. At least one node in the family graph network is a mobile phone number. The edges are used to indicate that the two connected nodes belong to the same family circle. The weight of the edge is associated with the family circle identification method that identifies the two nodes connected by the edge as belonging to the same family circle. The process iteratively executes the family graph network adjustment step. After completing this adjustment step, the information entropy of the family graph network no longer changes. The adjusted family graph network is then taken as the final family graph network, and nodes belonging to the same family circle in the final family graph network are considered as family members of one family. The family graph network adjustment step includes: The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative. Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

2. The method according to claim 1, characterized in that, After obtaining the corresponding feature data from the deep packet inspection data of each bandwidth, the method further includes: Based on the deep packet inspection data, mobile phone numbers that have connected to broadband more than a predetermined number of times within a predetermined time period will be removed.

3. The method according to claim 1, characterized in that, After identifying mobile phone numbers corresponding to the same grid space address as the third type of family circle based on grid space data, the method further includes: Extract family relationship data and call feature data corresponding to mobile phone numbers from the first type of family circle, the second type of family circle and the third type of family circle. The call feature data includes feature values ​​corresponding to multiple call features respectively. A deep neural network model is trained based on the call feature data to obtain a family member identification model, and family member identification is performed based on the family member identification model to obtain mobile phone number pairs; Adjust the first type of family circle, the second type of family circle, and the third type of family circle based on the mobile phone number.

4. The method according to claim 3, characterized in that, Before extracting call feature data of mobile phone numbers in the first type of family circle, the second type of family circle, and the third type of family circle, the method further includes: Based on the signaling data of the mobile phone number and the broadband installation address, determine the community where each mobile phone number is usually located at night; The extraction of family relationship data and call feature data corresponding to mobile phone numbers from the first type of family circle, the second type of family circle, and the third type of family circle includes: Extract family relationship data of mobile phone numbers in the first type of family circle, the second type of family circle, and the third type of family circle, and adjust the family relationship data according to the inconsistency of the nighttime residential communities corresponding to mobile phone numbers in the same family circle; Extract call feature data corresponding to each mobile phone number.

5. The method according to claim 4, characterized in that, The step of determining mobile phone numbers connected to the same home broadband network based on the deep packet inspection data, as the second type of home network, includes: Based on the deep packet inspection data, the mobile phone number connected to the home broadband is determined to be a mobile phone number from another network. The third-party application identifier in the deep packet inspection data and the mobile phone number connected to the home broadband are used as the second type of home circle. The method further includes: Based on the Wi-Fi location corresponding to the third-party application identifier in the deep packet inspection data, identify the nighttime resident cell corresponding to the third-party application identifier.

6. The method according to claim 4 or 5, characterized in that, The method further includes: The weights of edges between nodes in the family graph network are adjusted based on the inconsistency in the nighttime residential communities corresponding to mobile phone numbers within the same family circle.

7. A family member identification device, characterized in that, The device includes: The acquisition module is configured to retrieve data associated with mobile phone numbers from multiple data sources. The identification module is configured to identify telecommunications service-related data according to predetermined business rules to obtain a first type of household circle; obtain corresponding feature data from the deep packet inspection data of each broadband, the feature data including feature values ​​corresponding to pre-constructed features; perform clustering processing on the feature data of each broadband, and determine the household broadband from each broadband according to the clustering results; determine the mobile phone numbers connected to the same household broadband according to the deep packet inspection data, as the second type of household circle; and determine the mobile phone numbers corresponding to the same grid space address as the third type of household circle according to grid space data, with each type of household circle including at least one mobile phone number. The module is configured to build a family graph network based on various family circles. The family graph network includes nodes and edges connecting two nodes. At least one node in the family graph network is a mobile phone number. The edges are used to indicate that the two connected nodes belong to the same family circle. The weight of the edge is associated with a family circle identification method that identifies that the two nodes connected by the edge belong to the same family circle. The adjustment module is configured to iteratively execute the family graph network adjustment steps. After completing this adjustment step, the information entropy of the family graph network no longer changes. The adjusted family graph network is then taken as the final family graph network, and nodes belonging to the same family circle in the final family graph network are considered as family members of one family. The family graph network adjustment steps include: The family circle selection step is iteratively executed based on the family graph network until the family circles to which all nodes in the family graph network belong no longer change. The family circle selection step includes: for each node, sequentially attempting to add the node to each of the family circles adjacent to it, and for each family circle adjacent to the node, determining the information gain value when adding the node to that family circle. Specifically, when the family circle selection step is executed for the first time, each node in the family graph network is treated as a family circle; the maximum information gain value among the information gain values ​​is determined; and the node is added to the family circle corresponding to the maximum information gain value, based on the fact that the maximum information gain value is non-negative. Determine the information entropy of the family graph network, and determine whether the information entropy of the family graph network no longer changes based on the information entropy of the family graph network.

8. A computer-readable program medium, characterized in that, It stores computer program instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 6.

Images