Left-home family identification method and device, equipment and storage medium

Abstract

The application relates to the technical field of communication, and discloses a left-behind family identification method, device and equipment and a storage medium, the method comprising the following steps: identifying an old-young family based on a local user's social circle, identifying off-site workers through an off-site worker group identification model, determining the home of the off-site workers, and identifying a left-behind family according to the old-young family and the home of the off-site workers; since the off-site workers are identified through the off-site worker group identification model, and the left-behind family is identified according to the family member definition of the left-behind family, the old-young family is identified, the off-site workers are identified, and finally the off-site workers are associated with the old-young family, so that the left-behind family is identified, the accuracy of the off-site worker identification is improved, and the left-behind family identification is realized.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, device, and storage medium for identifying left-behind families. Background Technology

[0002] Current technologies extract return-to-hometown features solely from user call behavior, resulting in insufficient accuracy in prediction. Furthermore, there are no existing technical solutions for identifying families left behind in rural areas. Summary of the Invention

[0003] The main objective of this invention is to provide a method, apparatus, device, and storage medium for identifying left-behind families, aiming to solve the problem that existing technologies only extract return-to-hometown features from user call behavior, resulting in insufficient accuracy in prediction. Furthermore, existing technologies lack relevant technical solutions for identifying left-behind families.

[0004] To achieve the above objectives, the present invention provides a method for identifying left-behind families, the method comprising:

[0005] Identify elderly and young families based on local users' social circles;

[0006] The migrant worker group identification model is used to identify migrant workers and determine their family affiliation.

[0007] Left-behind families are identified based on the families of the elderly and children and the families to which the migrant workers belong.

[0008] Optionally, the step of identifying migrant workers through a migrant worker group identification model includes:

[0009] A wide data table is constructed based on the user's historical behavior within a preset time period. The wide data table includes at least one of the following: communication behavior information of the communication domain, location trajectory information of the service domain, family attribute information of the management domain, and family behavior information of the family domain.

[0010] Strongly correlated features and negatively correlated features are selected from the wide data table as inputs to the migrant worker return-to-hometown model, and migrant workers are identified through the migrant worker group identification model, which is constructed based on the decision tree algorithm.

[0011] Optionally, the method for identifying left-behind families further includes:

[0012] Obtain real-time location signaling triggered by users returning to their hometowns in other locations;

[0013] The migrant worker group identification model is updated based on the real-time location signaling.

[0014] Optionally, the method for identifying left-behind families further includes:

[0015] Obtain key business processing information of local family members and query the corresponding phone number information for the key business processing information;

[0016] The system identifies families left behind by matching the phone numbers of migrant workers with the phone numbers corresponding to the key business processing information.

[0017] Optionally, determining the family affiliation of the migrant worker includes:

[0018] Obtain the local close social circles of the aforementioned migrant workers;

[0019] The migrant worker's family affiliation is determined based on their local permanent residence and their local close social circle.

[0020] Optionally, the identification of elderly and young families based on local users' social circles includes:

[0021] Family groups are built based on local users' social circles, and location attributes are added to the family groups;

[0022] Identify elderly and young families based on family groups with added location attributes.

[0023] Optionally, the method for identifying left-behind families further includes:

[0024] Obtain all network relationship pairs, and select positive and negative samples based on the network relationship pairs;

[0025] Obtain the important and secondary features of the relationships between the entire network;

[0026] A social circle relationship recognition model is trained based on the positive samples, the negative samples, the important features, and the secondary features.

[0027] Based on the social circles established after training, the recognition model predicts the social circles of local users.

[0028] Furthermore, to achieve the above objectives, the present invention also proposes a device for identifying left-behind families, the device comprising:

[0029] The elderly and young family identification module is used to identify elderly and young families based on the local user's social circle.

[0030] The migrant worker identification module is used to identify migrant workers through a migrant worker group identification model and determine the family to which the migrant workers belong.

[0031] The left-behind family identification module is used to identify left-behind families based on the families of the elderly and children and the families to which the migrant workers belong.

[0032] In addition, to achieve the above objectives, the present invention also proposes a left-behind family identification device, which includes a memory, a processor, and a left-behind family identification program stored in the memory and executable on the processor. The left-behind family identification program is configured to implement the left-behind family identification method as described above.

[0033] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing a left-behind family identification program, which, when executed by a processor, implements the left-behind family identification method as described above.

[0034] This invention discloses a method for identifying elderly and young families based on local users' social circles, identifying migrant workers through a migrant worker group identification model, determining the families to which the migrant workers belong, and identifying left-behind families based on the elderly and young families and the families to which the migrant workers belong. Because this invention identifies migrant workers through a migrant worker group identification model, and sequentially identifies elderly and young families and migrant workers based on the definition of family members in left-behind families, and finally associates migrant workers with elderly and young families to identify left-behind families, it can improve the accuracy of migrant worker identification and achieve the identification of left-behind families. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the structure of a left-behind family identification device in the hardware operating environment involved in the embodiments of the present invention;

[0036] Figure 2 This is a flowchart illustrating the first embodiment of the method for identifying left-behind families according to the present invention;

[0037] Figure 3 This is a flowchart illustrating the second embodiment of the method for identifying left-behind families according to the present invention;

[0038] Figure 4 This is a schematic diagram illustrating the updating of the migrant worker group identification model according to an embodiment of the left-behind family identification method of the present invention;

[0039] Figure 5 This is a flowchart illustrating the third embodiment of the method for identifying left-behind families according to the present invention;

[0040] Figure 6 This is a structural block diagram of the first embodiment of the home-based family identification device of the present invention.

[0041] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0042] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0043] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a left-behind family identification device in the hardware operating environment of an embodiment of the present invention.

[0044] like Figure 1 As shown, the home-stay identification device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen, and optionally, it may also include a standard wired interface or a wireless interface. In this invention, the wired interface of the user interface 1003 may be a USB interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0045] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the identification device for left-behind families and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0046] like Figure 1 As shown, the memory 1005, which is identified as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a program for identifying left-behind families.

[0047] exist Figure 1 In the left-behind family identification device shown, the network interface 1004 is mainly used to connect to the back-end server and communicate with the back-end server; the user interface 1003 is mainly used to connect to the user equipment; the left-behind family identification device calls the left-behind family identification program stored in the memory 1005 through the processor 1001 and executes the left-behind family identification method provided in the embodiment of the present invention.

[0048] Based on the above hardware structure, an embodiment of the method for identifying left-behind families of the present invention is proposed.

[0049] Reference Figure 2 , Figure 2This is a flowchart illustrating the first embodiment of the method for identifying left-behind families according to the present invention, which presents the first embodiment of the method for identifying left-behind families according to the present invention.

[0050] It should be understood that existing technologies include predicting the return-to-hometown status of rural users based on the random forest algorithm. That is, using the random forest algorithm, the user's call behavior is used as a feature to predict whether the user will return home for the Spring Festival: a wide table is built based on the call behavior information of the target user in the previous months, and then the call behavior characteristics of the returning user are explored. First, the call behavior characteristics of the target user in October, November and December of year T-1 are used, and positive and negative samples are determined by whether the user returned home for the Spring Festival in year T, and a better model is selected. Then, this model is used to predict the call characteristics of the target user in August, September and October of year T. The following dimensions are mainly selected: (1) Basic user attribute information: local network information, package name, network of ownership, payment type, etc.; (2) User social circle information: social numbers, local network of social numbers, etc.; (3) User call information: call duration, number of calls, number of calls during holidays, number of outgoing calls, etc.

[0051] However, existing technologies extract return-to-hometown features solely from user call behavior, resulting in insufficient accuracy in prediction. Furthermore, there are no existing technical solutions for identifying families left behind in rural areas.

[0052] Therefore, in order to overcome the above-mentioned defects, this embodiment identifies migrant workers through a migrant worker group identification model, and identifies elderly and young families, migrant workers, and finally associates migrant workers with elderly and young families to identify left-behind families. This can improve the accuracy of migrant worker identification and enable the identification of left-behind families.

[0053] In the first embodiment, the method for identifying left-behind families includes:

[0054] Step S10: Identify elderly and young families based on local users' social circles.

[0055] It is understood that the execution subject of this embodiment may be a home-based identification device with data processing, network communication and program running functions, such as a computer, or other electronic devices that can achieve the same or similar functions. This embodiment does not limit this.

[0056] It should be noted that the social circle of a local user can be composed of family relationship pairs, and the elderly and children's family can be elderly people and / or children's families. This embodiment does not limit this.

[0057] It should be understood that identifying elderly and young families based on local users' social circles can specifically involve building family groups based on local users' social circles, adding location attributes to family groups, and identifying elderly and young families based on family groups with added location attributes.

[0058] Step S20: Identify migrant workers using the migrant worker group identification model and determine the family to which the migrant workers belong.

[0059] It should be noted that the identification model for migrant worker groups can be pre-built. For example, the identification model for migrant worker groups can be pre-built based on the decision tree algorithm.

[0060] Understandably, determining a migrant worker's family affiliation can be done by obtaining information about the migrant worker's close local social circle, and then determining the migrant worker's family affiliation based on the migrant worker's local permanent residence and close local social circle.

[0061] Step S30: Identify left-behind families based on the families of the elderly and children and the families to which the migrant workers belong.

[0062] In specific implementations, for example, identifying left-behind families based on the families to which the elderly and children belong and the families to which the migrant workers belong can be done by determining whether the migrant workers belong to families to which the elderly and children belong, and if so, then determining them as left-behind families.

[0063] Furthermore, to address the issue of family users (users belonging to the same household) among users from other provinces, the method for identifying left-behind families also includes:

[0064] Obtain key business processing information of local family members and query the corresponding number information of the key business processing information; identify left-behind families based on the migrant worker's number and the corresponding number information of the key business processing information.

[0065] Understandably, the system determines whether family members in the province (i.e., local family members) have subscribed to the mobile security service "Home Security Smart Monitoring" and whether the migrant worker's number is a service-bound or device-shared number. If the service relationship is currently valid, it indicates a family left behind.

[0066] When subscribing to the smart monitoring service, out-of-province phone numbers can be added to groups with local phone numbers. Parents working away from home can remotely view their elderly relatives and children at home via video. User A's family belongs to the same monitoring group member C. When A belongs to multiple monitoring groups, A is classified as a left-behind family if their permanent residence area is the same as the area A resided in as identified in step three.

[0067] This embodiment discloses a method for identifying elderly and young families based on local users' social circles, identifying migrant workers through a migrant worker group identification model, determining the families to which the migrant workers belong, and identifying left-behind families based on the elderly and young families and the families to which the migrant workers belong. Because this embodiment identifies migrant workers through a migrant worker group identification model, and identifies elderly and young families and migrant workers in sequence according to the definition of family members of left-behind families, and finally associates migrant workers with elderly and young families to identify left-behind families, it can improve the accuracy of migrant worker identification and achieve the identification of left-behind families.

[0068] Reference Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the left-behind family identification method of the present invention, based on the above. Figure 2 The first embodiment shown presents a second embodiment of the method for identifying left-behind families according to the present invention.

[0069] In the second embodiment, step S20 includes:

[0070] Step S201: Construct a wide data table based on the user's historical behavior within a preset time period. The wide data table includes at least one of the following: communication behavior information of the communication domain, location trajectory information of the service domain, family attribute information of the management domain, and family behavior information of the family domain.

[0071] It should be understood that, in order to improve the accuracy of the identification model for migrant worker groups, this embodiment adds a wide data table including communication behavior information in the communication domain, location trajectory information in the business domain, family attribute information in the management domain, and family behavior information in the family domain, as well as inputs for strong correlation features and negative correlation features.

[0072] Step S202: Select strongly correlated features and negatively correlated features from the wide data table as inputs to the migrant worker return home model, and identify migrant workers through the migrant worker group identification model, which is constructed based on the decision tree algorithm.

[0073] For ease of understanding, the following example illustrates the concept, but does not limit the scope of this solution. In one example, identifying migrant workers using a migrant worker group identification model includes the following steps:

[0074] 1. Construct a wide data table based on three consecutive years of user historical behavior:

[0075] For three consecutive years, during the Spring Festival period (from the Little New Year to the seventh day of the Lunar New Year), there were instances of people returning to their hometowns from other provinces, and these individuals had stable interactions with local residents in the previous year. Based on four major scenarios, we explored hypotheses to reconstruct the return-to-hometown scenario for migrant workers, accumulating the following user broad-based data categories: O-domain location trajectory information, including historical key migration trajectories and local permanent residence locations; B-domain communication behavior information, including communication circles, communication consumption, and service transactions; X-domain family behavior, including family broadband code parsing and terminal code parsing information; and complex family attributes, including a family group relationship identification model.

[0076] 2. Preprocess the data and select features.

[0077] 2.1 Output strong correlation features:

[0078] 2.1.1 Historical characteristics of returning home: Did you return home for the last Spring Festival, the Spring Festival before that, the Qingming Festival before that, and the Qingming Festival before that?

[0079] 2.1.2 Location Trajectory Characteristics: Landing and Permanent Location Information within the Province

[0080] 2.1.3 Create derived variables: family member consumption, family integration service processing

[0081] 2.2 Negative correlation characteristics: Those who return to their hometowns during winter and summer vacations (excluding students), those who have not communicated with users residing in the province for the past six months (excluding those with silent interactions), and those who joined the network within the past month (excluding users without a permanent location).

[0082] 3. Construct a decision tree algorithm for training and obtain the model's output results.

[0083] 3.1 Predicting Return Home: Number + Province

[0084] 3.2 Returning Home and Landing: Phone Number + Landing Area

[0085] 3.3 Returning Home and Staying: Phone Number + Area of ​​Stay + Duration

[0086] This embodiment adds a wide data table containing communication behavior information in the communication domain, location trajectory information in the service domain, family attribute information in the management domain, and family behavior information in the family domain, as well as inputs of strong correlation features and negative correlation features, thereby improving the accuracy of the migrant worker group identification model.

[0087] To further improve the accuracy of the migrant worker identification model, this embodiment also updates the migrant worker identification model based on real-time location signaling triggered by migrant users returning home. The method for identifying left-behind families also includes:

[0088] Obtain real-time location signaling triggered by users returning to their hometowns in other locations; update the migrant worker group identification model based on the real-time location signaling.

[0089] For ease of understanding, please refer to Figure 4 This explanation does not limit the scope of this solution. Figure 4 This is a schematic diagram of updating the migrant worker group identification model according to an embodiment of the method for identifying left-behind families of the present invention. In the diagram, when a migrant worker actually returns to his hometown and generates the first location signal within the province, his real-time location is input into the model. By querying a series of real-time location behaviors of the user, the data information on the return time and return location in the model is calibrated to improve the accuracy of the model.

[0090] In the second embodiment, step S20 further includes:

[0091] Step S203: Obtain the local close social circles of the migrant workers.

[0092] It should be understood that, in order to improve the accuracy of determining the family affiliation of migrant workers, this embodiment determines the family affiliation of migrant workers based on their local permanent residence and local close social circle.

[0093] Step S204: Determine the family to which the migrant worker belongs based on the migrant worker's local permanent residence and local close social circle.

[0094] For ease of understanding, the following example illustrates the concept, but does not limit the scope of this approach. In one example, determining the family affiliation of a migrant worker includes the following steps:

[0095] 1. Determine the close social circle of migrant workers within the province.

[0096] 1.1 Targeting the top 10 local users within the social circles of migrant workers.

[0097] The "Top 10 Local Users in the Social Circle" refers to all local users who have communicated with the user from another location, excluding non-personal user numbers and numbers with abnormal status. These users are then sorted in descending order of call duration, and the top 10 are selected. If fewer than 10 local users remain after excluding abnormal numbers, then the "Top 10 Local Users in the Social Circle" consists of all local users remaining after excluding abnormal numbers.

[0098] 1.2 Calculate the intimacy between users from other locations and the top 10 users in their social circles within the same province.

[0099] Based on the information of each user from the same province in the top 10 social circles of users from other regions, the intimacy f(ABi) between user A from other regions and each user Bi from this network is calculated using the following formula:

[0100] f(ABi) = min_max(number of calls between A and Bi in the current month) * 0.25 + min_max(call duration between A and Bi in the current month) * 0.25 + min_max(number of days between A and Bi in the current month) * 0.5

[0101] in:

[0102] min_max(Number of calls between A and Bi in the current month) = (Number of calls between A and Bi in the current month - Minimum number of calls in the current month) / (Maximum number of calls in the current month - Minimum number of calls in the current month)

[0103] min_max(Call duration between A and Bi in the current month) = (Call duration between A and Bi in the current month - Minimum call duration in the current month) / (Maximum call duration in the current month - Minimum call duration in the current month)

[0104] min_max(number of days A and Bi talked in the current month) = (number of days A and Bi talked in the current month - minimum number of days talked in the current month) / (maximum number of days talked in the current month - minimum number of days talked in the current month)

[0105] 2. Determining the family affiliation of migrant workers

[0106] 2.1 Based on the migrant worker's permanent residence location within the province, a distance filter is used to identify close friends / family members within the same province whose relationships match the geographical location (Bi). The maximum value of the intimacy score f(ABi) (B1) is output, representing the migrant worker's family members within the same province. The distance filter determines whether the "migrant worker's permanent residence location within the province" and the "permanent residence location of close friends / family members within the same neighborhood" fall within the same neighborhood location range.

[0107] 2.2 The maximum value of the intimacy rate f(ABi) is B1. The family in the province to which the migrant worker A belongs is the family to which the migrant worker A belongs, and thus the family to which the migrant worker A belongs is determined.

[0108] This embodiment determines the family affiliation of migrant workers based on their local permanent residence and close social circle, thereby improving the accuracy of family affiliation determination.

[0109] Reference Figure 5 , Figure 5 This is a flowchart illustrating the third embodiment of the method for identifying left-behind families according to the present invention. Based on the above embodiments, the third embodiment of the method for identifying left-behind families according to the present invention is proposed.

[0110] In the third embodiment, before step S10, the method further includes:

[0111] Step S01: Obtain all network relationship pairs and select positive and negative samples based on the network relationship pairs.

[0112] It should be understood that, in order to accurately predict the social circles of local users, in this embodiment, a social circle relationship pair recognition model is trained based on positive samples, negative samples, important features, and secondary features, and the social circle relationship pair recognition model is used to predict the social circles of local users.

[0113] Step S02: Obtain the important and secondary features between the network relationship pairs.

[0114] Step S03: Train the social circle relationship recognition model based on the positive samples, the negative samples, the important features, and the secondary features.

[0115] Step S04: Predict the social circles of local users based on the trained social circle relationships using the recognition model.

[0116] For ease of understanding, the following example illustrates the concept, but does not limit the scope of this solution. In one example, predicting a local user's social circle includes the following steps:

[0117] 1. Obtain all relationship pairs in the network and select the dataset:

[0118] All users who have had conversations with each other within the past 6 months form a set of network-wide relationship pairs.

[0119] Positive sample selection: Family groups that have applied for cloud family services, with each pair of users within a family labeled as label 1, and 1 million relationship pairs randomly selected from these groups as positive samples.

[0120] Negative sample selection: For family groups that have subscribed to the Cloud Family service, search for pairs of relationships between all users who have had conversations within the past 6 months. After removing pairs of relationships between users within the Cloud Family service, randomly select 1 million relationship pairs as negative samples.

[0121] At this point, 2 million positive and negative sample datasets have been extracted and divided into training, validation, and test sets according to a 6:2:2 ratio and a stratified extraction principle.

[0122] 2. Feature selection:

[0123] Select the following information between the relationship pairs as important characteristics: whether it is a secondary card, total call time, total number of calls, call time and number of calls between the caller and the recipient, call time and number of calls between the caller and the recipient at night, call time and number of calls between the caller and the recipient on weekends, call time and number of calls between the caller and the recipient during peak hours, number of night call frequency days and number of calls, number of weekend call frequency days and number of calls, and number of shared call contacts.

[0124] Select the following information between the relationship pairs as secondary features: caller / called number, caller / called gender, caller / called age, caller / called network duration, caller / called star rating, and caller / called gender.

[0125] 3. Model selection and training:

[0126] Feature vectors were constructed for the above features in the training set, and three classification methods—XGBoost, LightGBM, and Random Forest—were used for training, validation, and hyperparameter tuning.

[0127] The trained classification model is then applied to the test set data. After comparison and validation using metrics such as accuracy, recall, and AUC, the optimal classification model is selected.

[0128] 4. Model Prediction:

[0129] Prediction Target: After removing family user relationship pairs within the cloud family business from all relationship pairs across the entire network, and further removing negative sample sets, the remaining relationship pairs are used as the prediction target. The optimal classification model is used to predict them.

[0130] The predicted result of the relationship pair labeled 1, together with the family relationship pair formed by the cloud family business (where each pair of users within the family is labeled 1), forms the social circle relationship pair (i.e., the social circle of local users).

[0131] This embodiment trains a social circle relationship recognition model based on positive samples, negative samples, important features, and secondary features, and predicts the social circles of local users based on the trained social circle relationship recognition model, thereby accurately predicting the social circles of local users.

[0132] In the third embodiment, step S10 includes:

[0133] Step S101: Construct a family group based on the local user's social circle, and add a location attribute to the family group.

[0134] Understandably, in order to accurately identify families with elderly members and young children, this embodiment constructs family groups based on the social circles of local users and adds location attributes to the family groups, and identifies families with elderly members and young children based on the family groups with added location attributes.

[0135] It should be noted that "families with elderly and / or children" can refer to families with elderly people and / or children, and this embodiment does not impose any restrictions on this.

[0136] Step S102: Identify elderly and young families based on the family groups with added location attributes.

[0137] For ease of understanding, the following example is provided, but it does not limit the scope of this solution. In one example, identifying households with elderly members and young children includes the following steps:

[0138] 1. Define two numbers that are connected in family business as a relationship pair. Based on the relationship pairs, perform data fusion. If numbers in a set of data have family business relationships with each other, then define it as a family group.

[0139] The logic is as follows: a) Number 1 has business connections with other household numbers; b) Other numbers are interconnected. Specific algorithm:

[0140] Taking a three-person household as an example, a self-join is performed on the relationship pairs, generating number 1, number 2, and number 3, which are three related numbers. If number 1 appears twice after one self-join, it can be considered that this number is related to the other two-person households as a three-person household. If, after another self-join, number 1 appears twice squared, it indicates that this three-person household has no other redundant relationships, and number 2 and number 3 are related to each other. The priority of household business is shown in Table 1:

[0141] Table 1

[0142]

[0143]

[0144] 2. Add a location attribute to the group relationship obtained in step 1:

[0145] Iterate the family model based on family member scores Y:

[0146] Family member calculation rule: Y = ARPU * 0.8 / MAX(ARPU) + Network age * 0.2 / MAX(Network age)

[0147] ARPU and network tenure can be used to determine the scores of the top 5% of users. Values ​​above this value receive full marks, while values ​​below this value receive linear scores.

[0148] 2.1 Select family member A with the highest Y value as the central user for the first iteration.

[0149] 2.2 Compare with other family members. If the family member's nighttime residence is within 600 meters of the family member's residence, the family is grouped together. If the family member's nighttime residence is more than 600 meters away from the family member's residence, the family is not grouped together (family 1).

[0150] 2.3 After completing the first iteration, iterate over the remaining family members and select the family member B with the largest remaining user Y value as the center user for the second iteration.

[0151] 2.4 Compare with other family members. If the nighttime residence of member B is within 600 meters, they are grouped into the same family. If the nighttime residence of member B is greater than 600 meters, they are not grouped (family 2).

[0152] 2.5 Compare with other family members. If the nighttime residence of member C is within 600 meters, they are grouped into the same family. If the nighttime residence of member C is greater than 600 meters, they are not grouped (family 3).

[0153] 2.6 Other families aggregate into family 4.

[0154] 3. Identify households with elderly people and children.

[0155] For the two family models, we will determine whether they are families with elderly members or children. The criterion is that if at least one family member is identified as an elderly person or a child, the family can be considered a family with elderly members or children. The determination scheme is as follows, and only one condition needs to be met.

[0156] 3.1 ID Card Age: By registering phone card information using ID card, the user's age is analyzed. Numbers for males aged 60 and above or females aged 55 and above are for elderly households, and numbers for those aged 16 and below are for children households.

[0157] 3.2 Parsing user's device model information: If the user's terminal model is a children's device or a children's watch, it is identified as a family with children;

[0158] 3.3 DPI Analysis of Internet Applications for the Elderly and Children: A knowledge base for apps used by the elderly and children is created. Mobile internet DPI analysis data outputs information such as the number of days the app was used in the current month, the data usage in the current month, the usage time in minutes, and the number of times the app was opened. Based on the knowledge bases for apps used by the elderly and children, the data for the current month is summarized. Each of these data is multiplied by a weight and summed to obtain an app usage index. A threshold is set for the index; if it exceeds the threshold, the app is categorized and classified as belonging to an elderly or child-oriented household.

[0159] This embodiment constructs family groups based on local users' social circles and adds location attributes to these family groups. It then identifies families with elderly members and children based on the family groups with added location attributes, thereby enabling accurate identification of such families.

[0160] In addition, refer to Figure 6 This invention also proposes a device for identifying left-behind families, the device comprising:

[0161] The elderly and young family identification module 10 is used to identify elderly and young families based on the local user's social circle.

[0162] The migrant worker identification module 20 is used to identify migrant workers through a migrant worker group identification model and determine the family to which the migrant workers belong.

[0163] The left-behind family identification module 30 is used to identify left-behind families based on the families of the elderly and children and the families to which the migrant workers belong.

[0164] This embodiment discloses a method for identifying elderly and young families based on local users' social circles, identifying migrant workers through a migrant worker group identification model, determining the families to which the migrant workers belong, and identifying left-behind families based on the elderly and young families and the families to which the migrant workers belong. Because this embodiment identifies migrant workers through a migrant worker group identification model, and identifies elderly and young families and migrant workers in sequence according to the definition of family members of left-behind families, and finally associates migrant workers with elderly and young families to identify left-behind families, it can improve the accuracy of migrant worker identification and achieve the identification of left-behind families.

[0165] In one embodiment, the migrant worker identification module 20 is further configured to construct a wide data table based on the user's historical behavior within a preset time period. The wide data table includes at least one of the following: communication behavior information in the communication domain, location trajectory information in the business domain, family attribute information in the management domain, and family behavior information in the family domain. Strongly correlated features and negatively correlated features are selected from the wide data table as inputs to the migrant worker return-to-hometown model, and migrant workers are identified through the migrant worker group identification model, which is constructed based on a decision tree algorithm.

[0166] In one embodiment, the migrant worker identification module 20 is further configured to acquire real-time location signaling triggered by users returning to their hometowns; and update the migrant worker group identification model based on the real-time location signaling.

[0167] In one embodiment, the left-behind family identification module 30 is further configured to obtain key business handling information of local family members and query the number information corresponding to the key business handling information; and identify left-behind families based on the migrant worker's number and the number information corresponding to the key business handling information.

[0168] In one embodiment, the migrant worker identification module 20 is further configured to obtain the local close social circle of the migrant worker; and determine the migrant worker's family affiliation based on the migrant worker's local permanent residence and the local close social circle.

[0169] In one embodiment, the elderly-child family identification module 10 is further configured to construct family groups based on the social circles of local users and add location attributes to the family groups; and identify elderly-child families based on the family groups with added location attributes.

[0170] In one embodiment, the left-behind family identification device further includes:

[0171] The social circle prediction module is used to obtain all network relationship pairs and select positive and negative samples based on the network relationship pairs; obtain important and secondary features between the network relationship pairs; train a social circle relationship pair recognition model based on the positive samples, the negative samples, the important features, and the secondary features; and predict the social circle of local users based on the trained social circle relationship pair recognition model.

[0172] Other embodiments or specific implementations of the left-behind family identification device of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0173] Furthermore, this embodiment of the invention also proposes a storage medium storing a left-behind family identification program, which, when executed by a processor, implements the left-behind family identification method as described above.

[0174] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0175] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0176] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0177] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for identifying left-behind families, characterized in that, The method for identifying left-behind families includes: Identify elderly and young families based on local users' social circles; The migrant worker group identification model is used to identify migrant workers and determine their family affiliation. Left-behind families are identified based on the families of the elderly and children and the families to which the migrant workers belong; The method of identifying migrant workers through a migrant worker group identification model includes: A wide data table is constructed based on the user's historical behavior within a preset time period. The wide data table includes at least one of the following: communication behavior information of the communication domain, location trajectory information of the service domain, family attribute information of the management domain, and family behavior information of the family domain. Strongly correlated features and negatively correlated features are selected from the wide data table as inputs to the migrant worker return home model, and migrant workers are identified through the migrant worker group identification model, which is constructed based on the decision tree algorithm. The strongly correlated features are historical return home features, location trajectory features, and derived variables.

2. The method for identifying left-behind families as described in claim 1, characterized in that, The method for identifying left-behind families also includes: Obtain real-time location signaling triggered by users returning to their hometowns in other locations; The migrant worker group identification model is updated based on the real-time location signaling.

3. The method for identifying left-behind families as described in claim 1, characterized in that, The method for identifying left-behind families also includes: Obtain key business processing information of local family members and query the corresponding phone number information for the key business processing information; The system identifies families left behind by matching the phone numbers of migrant workers with the phone numbers corresponding to the key business processing information.

4. The method for identifying left-behind families as described in any one of claims 1 to 3, characterized in that, The determination of the family affiliation of the migrant workers includes: Obtain the local close social circles of the aforementioned migrant workers; The migrant worker's family affiliation is determined based on their local permanent residence and their local close social circle.

5. The method for identifying left-behind families as described in any one of claims 1 to 3, characterized in that, The method of identifying elderly and young families based on local users' social circles includes: Family groups are built based on local users' social circles, and location attributes are added to the family groups; Identify elderly and young families based on family groups with added location attributes.

6. The method for identifying left-behind families as described in any one of claims 1 to 3, characterized in that, The method for identifying left-behind families also includes: Obtain all network relationship pairs, and select positive and negative samples based on the network relationship pairs; Obtain the important and secondary features of the relationships between the entire network; A social circle relationship recognition model is trained based on the positive samples, the negative samples, the important features, and the secondary features. Based on the social circles established after training, the recognition model predicts the social circles of local users.

7. A device for identifying left-behind families, characterized in that, The device for identifying left-behind families includes: The elderly and young family identification module is used to identify elderly and young families based on the local user's social circle. The migrant worker identification module is used to identify migrant workers through a migrant worker group identification model and determine the family to which the migrant workers belong. The left-behind family identification module is used to identify left-behind families based on the families of the elderly and children and the families to which the migrant workers belong; The migrant worker identification module is further configured to construct a wide data table based on users' historical behavior within a preset time period. The wide data table includes at least one of the following: communication behavior information in the communication domain, location trajectory information in the business domain, family attribute information in the management domain, and family behavior information in the family domain. Strongly correlated features and negatively correlated features are selected from the wide data table as inputs to the migrant worker return-home model, and migrant workers are identified through the migrant worker group identification model. The migrant worker group identification model is constructed based on a decision tree algorithm, and the strongly correlated features are historical return-home features, location trajectory features, and derived variables.

8. A device for identifying left-behind families, characterized in that, The device for identifying left-behind families includes: a memory, a processor, and a program for identifying left-behind families stored in the memory and executable on the processor. When the program for identifying left-behind families is executed by the processor, it implements the method for identifying left-behind families as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores a program for identifying left-behind families, which, when executed by a processor, implements the method for identifying left-behind families as described in any one of claims 1 to 6.

Images