Method and apparatus for positioning an out-of-network user and electronic device

Abstract

The present disclosure relates to a method and device for positioning out-of-network users and an electronic device, and relates to the technical field of communication networks. The method comprises: counting communication signaling, and communication behavior data, location data and working time characteristics of in-network users in a specific time interval; filtering in-network users using the communication signaling, the communication behavior data, the location data and the working time characteristics to obtain target in-network users with different logistics occupations in the specific time interval; filtering candidate base stations where the target in-network users and out-of-network users come into contact; and inputting cumulative communication behavior data of the out-of-network users and the target in-network users at the candidate base stations into a trained out-of-network user residence model to obtain the residence of the out-of-network users in different time periods. In this way, the problem that the social circle cannot be effectively extracted and the residence of the out-of-network users cannot be accurately calculated can be avoided, and the residence of the out-of-network users can be accurately positioned.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication networks, and in particular to a method and device for positioning out-of-network users and an electronic device. BACKGROUND

[0002] In the current environment of fierce competition in the communication market, in order to improve the efficiency of customer acquisition and retention, the marketing and support teams of operators have long been faced with the challenge of how to effectively identify and contact potential customers. In particular, the mining of out-of-network users, i.e. attracting potential users from competitors, has become a key issue.

[0003] Currently, some operators have begun to try to use technical means to analyze the social circles of in-network customers and out-of-network customers, to calculate whether the out-of-network customers are in the same unit or the same family circle as the in-network customers, and to calculate the daytime and nighttime residence of the out-of-network customers through the residence mode of the in-network customers.

[0004] However, in the existing technology, the effect of calculating the residence of the out-of-network customers through the social circle is affected by the social circle, but in the case where the social circle is not easy to identify, for example, the social circle of employees in small and medium-sized shops is relatively simple, and it is difficult to effectively calculate the residence of the out-of-network customers. SUMMARY

[0005] Therefore, the present application provides a method and device for positioning out-of-network users and an electronic device, which mainly aims to solve the technical problem that it is difficult to effectively calculate the residence of out-of-network customers in the case where the social circle is not easy to identify in the process of identifying the residence of out-of-network customers.

[0006] According to a first aspect of the present disclosure, a method for positioning out-of-network users is provided, which comprises:

[0007] statistically analyzing communication signaling, and communication behavior data, location data and working time characteristics of in-network users in a specific time interval;

[0008] screening the in-network users using the communication signaling, the communication behavior data, the location data and the working time characteristics, to screen out target in-network users with different logistics occupations in the specific time interval;

[0009] screening candidate base stations where the target in-network users contact out-of-network users;

[0010] inputting cumulative communication behavior data of the out-of-network users and the target in-network users at the candidate base stations into a trained out-of-network user residence model, to obtain the residence of the out-of-network users in different time periods.

[0011] According to a second aspect of the present disclosure, there is provided a device for positioning out-of-network users, comprising:

[0012] a statistics module configured to count communication signaling, and communication behavior data, location data and working time characteristics of in-network users in a specific time interval;

[0013] a first screening module configured to screen the in-network users using the communication signaling, the communication behavior data, the location data and the working time characteristics, and screen target in-network users with different logistics professions in the specific time interval;

[0014] a second screening module configured to screen candidate base stations where the target in-network users contact out-of-network users;

[0015] an input module configured to input cumulative communication behavior data of the out-of-network users and the target in-network users at the candidate base stations into a trained out-of-network user residence model, and obtain a residence of the out-of-network users in different time periods.

[0016] According to a third aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of the first aspect.

[0017] According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are configured to cause a computer to perform the method of the first aspect.

[0018] According to a fifth aspect of the present disclosure, there is provided a computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the method of the first aspect.

[0019] The method, device and electronic equipment for positioning an out-of-network user provided by the present disclosure are compared with the prior art. The present disclosure statistically analyzes communication signaling, communication behavior data, location data and working time characteristics of in-network users in a specific time interval. The in-network users are filtered by using the communication signaling, communication behavior data, location data and working time characteristics, and target in-network users with different logistics occupations in the specific time interval are filtered. Candidate base stations where the target in-network users and the out-of-network user come into contact are filtered. The cumulative communication behavior data of the out-of-network user and the target in-network user at the candidate base stations are input into a trained out-of-network user permanent residence model, and the permanent residence of the out-of-network user in different time periods is obtained. In order to avoid the case that the social circle cannot be effectively extracted, resulting in the inability to accurately calculate the permanent residence of the out-of-network user, the target in-network users with different logistics occupations are filtered in this way, and the communication behavior data of the target in-network user and the out-of-network user at different contact base stations (i.e., candidate base stations) in different time periods are used for judgment, so as to accurately position the permanent residence of the out-of-network user.

[0020] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings incorporated in the specification and forming a part thereof illustrate embodiments consistent with the present application and together with the description serve to explain the principles of the present application.

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the technical solutions of the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0023] Figure 1 A flowchart of a method for positioning an out-of-network user provided by an embodiment of the present disclosure;

[0024] Figure 2 A flowchart of a method for positioning an out-of-network user provided by an embodiment of the present disclosure;

[0025] Figure 3 A flowchart of a method for identifying the permanent residence of an out-of-network customer provided by an embodiment of the present disclosure;

[0026] Figure 4 A logistics user occupation feature map provided by an embodiment of the present disclosure;

[0027] Figure 5This is a candidate base station diagram provided in the embodiments of this disclosure;

[0028] Figure 6 This is a cross-network permanent residence identification map based on multiple occupational characteristics time series provided in the embodiments of this disclosure;

[0029] Figure 7 This is a diagram of a cross-network customer permanent residence identification system provided in an embodiment of the present disclosure;

[0030] Figure 8 This is a schematic diagram of the structure of a cross-network user positioning device provided in an embodiment of this disclosure. Detailed Implementation

[0031] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments of this disclosure and the features described therein can be combined with each other.

[0032] The following description, with reference to the accompanying drawings, outlines a method, apparatus, and electronic device for locating users on different networks according to embodiments of this disclosure.

[0033] This disclosure provides a method, device, and electronic device for locating users on different networks. By screening target users on the local network with different logistics occupations, and by distinguishing between target users on the local network and users on different networks under different contact base stations (i.e., candidate base stations) in different time periods, the permanent residence of users on different networks can be accurately located.

[0034] like Figure 1 As shown, embodiments of this disclosure provide a method for locating users on a different network, the method including:

[0035] Step 101: Collect statistical communication signaling data, as well as communication behavior data, location data, and working time characteristics of users on this network within a specific time interval.

[0036] Among them, communication signaling can be signals generated by communication operators during communication, used to reflect the activities of users in the communication network; communication behavior data can include the number of calls, call duration, number of base stations, etc. of users in a specific time interval; location data can be obtained through base station positioning technology, used to determine the location of users in different time points.

[0037] Working time characteristics can include users' activity patterns and working hours within a specific time interval. For example, food delivery workers and couriers may work during certain time periods, while ride-hailing drivers may be active during other time periods. Analyzing working time characteristics can help identify the occupations of users on this network more effectively.

[0038] Step 102: Filter users of this network using communication signaling, communication behavior data, location data, and working time characteristics to identify target users of this network with different logistics occupations within a specific time interval.

[0039] Among them, the users of this network can be end users who use a specific communication network (such as mobile communication network); the target users of this network can be users of this network with different logistics occupations, such as food delivery workers, couriers, and Didi drivers of this network.

[0040] In this embodiment of the disclosure, target users of the local network who engage in different logistics professions (such as couriers, freight drivers, etc.) within a specific time interval are identified and filtered through communication signaling, communication behavior data (such as telephone call records, SMS records, internet browsing behavior, etc.), location data, and working time characteristics. Communication behavior data between these target users and users on other networks is then mined. This data is then compared with communication behavior data of the target users at different contact base stations within different time periods. In this way, the base station at the location of the user on another network can be accurately identified. Finally, combined with local network base station positioning technology, the location of the user on another network can be accurately determined.

[0041] Step 103: Filter candidate base stations for contact between target users on this network and users on other networks.

[0042] Among them, candidate base stations can be multiple base stations that may exist when locating the user's permanent residence, which can be used to locate the user's workplace, home address or other frequently visited places on other networks.

[0043] One possible approach is to extract communication behavior data between the target user on the local network and users on other networks, including call logs, SMS records, and network usage records. This communication behavior data can be used to determine the base stations the target user and the user on other networks connected to during their communication. By analyzing the user's communication signaling data, the connection relationships between base stations can be captured; that is, a base station relationship network can be constructed through base station handover. Within this network, by analyzing the base station handover events during the communication between the target user and the user on other networks, a series of base stations that may have been involved in the communication process can be identified—these are candidate base stations.

[0044] Step 104: Input the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods.

[0045] Among them, the model of the permanent residence of users on different networks may include, but is not limited to, the triple LSTM model, which can be a model that combines three long short-term memory networks (LSTM).

[0046] In this embodiment of the disclosure, a triple LSTM model can be used to extract communication behavior features from the cumulative communication behavior data between users on different networks and users on the target local network. The triple LSTM model can determine the communication behavior between users on different networks and users on the target local network by analyzing the communication behavior features. Communication behavior features may include call duration, frequency, time distribution, changes in communication base stations, etc. By analyzing the communication behavior, the signaling data of the target local network user contacting the base station within a specific time period can be more accurately identified, and the permanent residence location can be extracted as the permanent residence location of the user on the different network.

[0047] In summary, compared with existing technologies, the cross-network user location method provided in this disclosure utilizes statistical communication signaling, as well as communication behavior data, location data, and working time characteristics of users within a specific time interval. It then filters users within the same network using these data to identify target users with different logistics occupations within that specific time interval. Next, it filters candidate base stations where target users and cross-network users interact. Finally, it inputs the cumulative communication behavior data of cross-network users and target users at these candidate base stations into a trained cross-network user location model to obtain the cross-network user's permanent residence location within different time periods. To avoid situations where the interaction circle cannot be effectively extracted, leading to inaccurate estimation of cross-network user locations, this method filters target users with different logistics occupations and uses communication behavior data of target users and cross-network users at different contact base stations (i.e., candidate base stations) within different time periods to accurately locate the cross-network user's permanent residence.

[0048] Furthermore, as a refinement and extension of the above embodiments, and in order to fully illustrate the specific implementation process of the method disclosed herein, this disclosure provides the following... Figure 2 The specific method shown includes:

[0049] Step 201: Collect statistical communication signaling data, as well as communication behavior data, location data, and working time characteristics of users on this network within a specific time interval.

[0050] For embodiments of this disclosure, such as Figure 3 As shown, the flowchart for identifying permanent resident communities includes the following main modules:

[0051] User communication behavior extraction module: This module mainly extracts data such as the number of calls, duration, number of base stations, number of visits to related apps, and traffic from the call records of the user's residence at certain time intervals.

[0052] Base station extraction module: This module mainly constructs a relationship network of base stations in the local network to filter out the effective base station groups that have contact with logistics personnel for each user and identify them as candidate base stations.

[0053] Occupational type identification module: This module mainly uses classic machine learning algorithms to identify occupational users such as logistics workers based on data from telecommunications operators, including location and call records, thereby reducing unnecessary model screening overhead and improving accuracy.

[0054] Occupational Feature Map Construction Module: This module mainly constructs a 4-dimensional feature map based on the feature information extracted above, with 30-minute intervals per day and information from the past 30 days, to describe the trajectory characteristics of users in the logistics industry on this network.

[0055] Occupational Feature Recognition Module: This module primarily employs deep learning strategies, combining feature map input data from the three types of occupations mentioned above to train a relevant convolutional network. Based on this network, it determines the user's occupation and ultimately identifies which logistics company the user works for by considering factors such as the apps they visit and their social circles.

[0056] Trajectory Feature Map Construction Module: This module mainly uses data such as communication behavior of users from other networks and the aforementioned logistics professionals extracted from the local network under base stations and at different time periods to construct occupational feature trajectory map data for different logistics industries.

[0057] Cross-network resident user model training module: This module uses a triple-fusion LSTM algorithm to train the model on the occupational feature map data of the aforementioned logistics occupational users on this network. By using positive and negative sample data of users on this network, it extracts model features of cross-network users at different time periods and different base stations, comparing them with users of different logistics occupational types on this network.

[0058] Inter-network permanent residence identification module: This module determines whether a base station is the permanent residence of an abnormal user by inputting the time period, number of calls, and duration of contact between an inter-network user and a logistics user of the same type under different contact types on the local network at the base station.

[0059] Step 202: Use communication signaling, communication behavior data, location data, and working time characteristics to filter users on this network and identify users with specific occupational characteristics within a specific time interval.

[0060] In this embodiment of the disclosure, firstly, communication signaling, communication behavior data, location data, and working time characteristics can be used to initially screen users of the network, thereby initially determining whether users of the network have the characteristics of a logistics profession, which helps to narrow down the identification range.

[0061] Based on the initial screening, we can further combine the professional habits of our users to more accurately identify whether they belong to the logistics profession. Professional habits can include the user's behavior patterns and activity routines at work.

[0062] In specific application scenarios, identifying logistics professions by app name alone cannot comprehensively collect all relevant app names used by logistics professions such as express delivery, food delivery, and ride-hailing. Therefore, simply identifying by app name is not an effective way to achieve accurate classification. Since the occupational characteristics of the above three logistics professions can be initially screened using signaling data, a method based on communication operator signaling and communication data is used to distinguish these three professions from other professions. The main fields used are: number of calling users (daily), average call duration (daily), number of local base stations (daily), average local base station dwell time (daily), and the duration of frequently used app addresses after excluding commonly used social apps. Data is collected from 7:30 AM to 10:30 PM, Monday to Friday, excluding data on users' frequently used base stations. The XGBoost model is used to perform binary classification on the above three types plus other types. Positive samples are selected from users of mainstream apps on this website (JD.com and STO Express for express delivery; Ele.me and Meituan for food delivery; Didi and Kuaidi for ride-hailing services, etc.) based on the above 8 fields. Negative samples are selected from group user data of categories A, B, C, and D. The sample size is 1:100. Then, they are selected into the candidate group for occupational feature recognition.

[0063] Step 203: Construct an occupational characteristic map of users with specific occupational characteristics by using their communication behavior data, location data, and working time characteristics.

[0064] In order to accurately identify logistics professionals such as food delivery workers, couriers, and ride-hailing drivers, this invention combines communication behavior data, location data, and the working time characteristics of the three types of logistics personnel to collect information such as the number of calls, duration, number of base stations, and number of times related apps were used by users of this network within a historical time period (e.g., the past month, with half-hour intervals) to construct a professional feature map of users of this network.

[0065] In specific application scenarios, such as Figure 4As shown, different professions have specific behavioral characteristics. For example, food delivery is concentrated during mealtimes, and couriers have fixed routes, so their delivery times to their assigned areas are relatively fixed. To accurately identify food delivery drivers, same-city couriers, logistics couriers from SF Express and JD.com, ride-hailing drivers, and other professional users, a four-dimensional feature map (X, Y, Z, W) is constructed based on the behavioral characteristics of professional users. Here, X represents the past 30 days for the aforementioned profession, Y represents the number of columns in a day (divided into 48 columns at 30-minute intervals), Z represents whether it is a holiday, and W represents the characteristics of the aforementioned professions. The feature map of this network user combines the above five characteristics in the X, Y, and Z dimensions with seven additional dimensions, including average daily stay at indoor distributed base stations and the number of different indoor distributed base stations stayed at per day. Therefore, the dimensions of this network user feature map are (30, 48, 2, 7). Local behavioral characteristics are extracted as relevant information about the aforementioned professions in their place of residence.

[0066] Step 204: Use a convolutional neural network algorithm to extract occupational features from the occupational feature map, and identify target users with different logistics occupations based on the occupational features.

[0067] In this embodiment of the disclosure, the occupational feature map can be input into a convolutional neural network algorithm. Through the hierarchical structure of the neural network, such as convolutional layers and pooling layers, occupational features in the occupational feature map can be extracted. In particular, to adapt to the CONV2D convolutional neural network, Z and W in the four-dimensional feature map are merged, and the dimensions become (X, Y, Z*W) three-dimensional.

[0068] Finally, the softmax function can be used to identify the above occupational characteristics, and the identification results are divided into three occupational characteristic categories: food deliveryman, same-city courier, logistics courier for SF Express, JD.com, and ride-hailing driver.

[0069] The convolutional network can be the classic VGG-11 network, with each convolutional kernel having a size of 3*3, and the number of channels from top to bottom being 64, 128, 256, 256, 512, 512, 512, 512, 4096, and 4, with a stride of 1. The activation function is ReLU, and the pooling layers use a 3*3 max pooling strategy, located after layers 1, 2, 4, 6, and 8, with a stride of 1.

[0070] Step 205: Construct a base station relationship network; use the base station relationship network to screen candidate base stations that allow users on the target network to interact with users on other networks.

[0071] Among them, "inter-network users" can refer to users who use network services from other operators within a mobile communication network. For example, if a user's mobile phone number is registered with a different operator than the network operator they are currently accessing, then that user is called an "inter-network user."

[0072] In this embodiment of the disclosure, to reduce the number of candidate base stations and more accurately pinpoint the base station closest to the user's permanent residence, while reducing unnecessary computational overhead, a base station relationship network is constructed to filter base stations near each user's permanent residence. Initial screening is performed on base stations used for communication between the user and logistics staff. Since multiple base stations may exist near the permanent residence, the DBSCAN algorithm is used to cluster the base stations and eliminate isolated points, such as... Figure 5 As shown.

[0073] Specifically, all appearing base stations can be extracted from the call records of users on different networks and logistics personnel, and these base stations can be used as preliminary candidate results for base station screening.

[0074] Secondly, a network of relationships between base stations can be established using signaling data from users within the network. This signaling data can reveal how users switch from one base station to another during their activity. In other words, a network of base station relationships is constructed through these handovers. If a user switches directly from base station A to base station B during their activity, then if there is a direct relationship between base station A and base station B, it forms an undirected edge. Through this undirected edge, the shortest path between base station A and base station B in the relationship network can be obtained. The number of edges traversed by the shortest path describes the distance between base station A and base station B.

[0075] Finally, the DBSCAN algorithm can be used to cluster the preliminary candidate base station relationships for each inter-network user and logistics personnel. The distance between base stations is the number of edges traversed by the shortest path in the relationship network described above, and non-resident base stations, i.e., points that do not belong to any cluster, are eliminated. The remaining base stations are candidate base stations for accurately locating the user's permanent residence.

[0076] DBSCAN is a density-based clustering algorithm that can discover clusters of arbitrary shapes in noisy spatial databases and separate the clusters from the noise.

[0077] In this embodiment of the disclosure, a base station relationship network is constructed; the base station relationship network is used to screen candidate base stations that the target local user and the user of the other network can contact. Specifically, this may include: determining multiple effective base stations that the target local user and the user of the other network can contact; determining the base station distance between multiple effective base stations using the base station relationship network; clustering multiple effective base stations according to the base station distance to obtain clustering results; and removing non-local base stations from the clustering results to obtain candidate base stations.

[0078] Among them, multiple effective base stations can be a group of base stations that have frequent contact between logistics personnel on this network and users on other networks, and can stably reflect the activity trajectory of logistics personnel within a certain period of time.

[0079] Step 206: Based on the communication behavior data between the target local network user and different base stations, identify the users from other networks; based on the communication behavior data between the users from other networks and the target local network user under the candidate base stations, construct the occupational characteristic trajectory map of the target local network user.

[0080] In this embodiment of the disclosure, a trajectory feature map construction module can be used to construct occupational feature trajectory map data for different logistics industries based on communication behavior data extracted from base stations on the local network and at different time periods for users from other networks and logistics professionals. The occupational feature trajectory map can be used to describe the trajectory characteristics of logistics industry users on the local network.

[0081] Step 207: Input the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods.

[0082] In this embodiment of the disclosure, by integrating behavioral data from three different levels of contact—such as food delivery workers and local couriers (close-contact type), logistics couriers (medium-contact type), and ride-hailing drivers (long-contact type)—a fusion temporal deep learning algorithm is used to extract the local network base station communication during the most likely time period, thereby identifying the final permanent location of users on different networks.

[0083] Training methods for the model of the permanent residence of users on different networks may include:

[0084] Extract historical trajectory features from the occupational feature trajectory map, as well as historical cumulative communication behavior data and historical permanent residence data of users from other networks and target users from the same network in different historical time periods under candidate base stations; use historical cumulative communication behavior data and historical trajectory features as input features, and use the historical permanent residence data of users from other networks in different historical time periods as training labels, and iteratively train the permanent residence model of users from other networks until the loss function is less than a preset threshold.

[0085] The preset threshold can be a value greater than 0 and less than 1. The closer the preset threshold is to 0, the higher the prediction accuracy of the model for the permanent residence of users on different networks. The specific value can be set according to the actual application scenario, and no specific limitation is made here.

[0086] Historical trajectory features can be defined as the cumulative number of calls and cumulative dwell time of a target user on the same base station within different historical time periods. Additionally, to reflect the characteristics of different logistics companies, a logistics company dimension is added. Therefore, historical trajectory features can be composed of four dimensions: (X, Y, Z, C, C1). X divides a day into 24 columns at 60-minute intervals; Y represents the 7 days from Monday to Sunday; Z represents the cumulative number of calls and cumulative call duration. Since food delivery and same-city express delivery are both short-distance contact types, they are combined into one. Therefore, C represents three types of logistics occupations: short, medium, and long-distance contact, such as food delivery, express delivery, and ride-hailing. C1 represents different logistics companies under these three different contact types, for example, food delivery includes logistics companies such as Meituan and Ele.me.

[0087] In this embodiment of the disclosure, the cumulative communication behavior data between the inter-network user and the target local network user under the candidate base station is input into the trained inter-network user's permanent residence model to obtain the inter-network user's permanent residence in different time periods, which may specifically include:

[0088] Calculate the weight coefficient of the probability of activity of the target network user under each candidate base station; use the attention mechanism and softmax function to process the weight coefficients to obtain the weight coefficients corresponding to different types of logistics occupations; concatenate the processed data of different types of logistics occupations and their corresponding weight coefficients, perform calculations through a fully connected layer, and use the softmax function to calculate whether the candidate base station is the permanent residence of the user from another network.

[0089] Specifically, to extract users' permanent residence locations through occupational characteristics, a multi-temporal series-based inter-network permanent residence location identification module, tri-lstm, was designed based on the aforementioned trajectory feature map. Positive samples of the module's input data can represent the contact information of different users with the aforementioned three types of logistics at their permanent residence's home base station during different times of the week (7 days a week). Negative samples can represent the contact information of users with the aforementioned three types of logistics at non-permanent residence's home base station during different times of the week (7 days a week). Figure 6 As shown.

[0090] First, an encoding model can be constructed. The encoding model can construct three encoder models according to three different contact types of the same user. Each encoder model adopts the LSTM algorithm, and the above X dimension has 24 columns. The LSTM is set to contain 24 steps.

[0091] Secondly, the input to each step of the encoder model can be the cumulative number of calls made by the same user at the same base station, and the cumulative dwell time [V1, V2]. To reflect the characteristics of different courier companies, the above cumulative values ​​are expanded into a two-dimensional matrix [n, 2] according to the courier company. To accurately capture the input parameters, three heads are designed, each with a weight W of [1, 16]. The input at each step is mapped to different spaces through multiple heads.

[0092] Then, based on the weight output of the LSTM model at each step, the attention mechanism can be fused to apply the state LSTM at each step. state *W attention Then, the weights W are assigned to each step using the softmax function. step And correct each step of LSTM output *W step .

[0093] Then, the output of the final step of the LSTM for the above three job types is weighted according to the job type focus mechanism, LSTM' output =lstm output *W′ attention The three types of lstm' after adjustment output The data is then spliced ​​together and calculated with the final fully connected layer. Finally, a softmax method is used to determine whether the base station is the user's permanent residence.

[0094] Finally, the trained model is used to determine whether a base station is a user's permanent residence by analyzing the communication behavior data of users from different networks at different distances from the local network and for different logistics types at different time periods.

[0095] The inter-network user permanent residence identification device can extract the communication behavior of users on the local network and use an effective logistics personnel identification model to identify users from different logistics companies under different contact types. By combining the movement trajectory map of logistics personnel, an inter-network user permanent residence identification model can be constructed, which can accurately identify the permanent residence of inter-network users based on the contact timing of logistics personnel on the local network.

[0096] like Figure 7 As shown, the main steps of this application are as follows:

[0097] First, data such as the number of calls, duration, number of base stations, and number of visits to related apps are extracted from the communication behavior of all users on this network at certain time intervals, generating traffic data.

[0098] Based on user communication data and signaling location data, machine learning algorithms are used to identify logistics and other professional users, providing a basis for the accurate identification of logistics personnel in the future.

[0099] Based on the logistics occupational characteristic map, a day is divided into 48 columns with 30-minute intervals. Z represents whether it is a holiday or not, and W represents the five dimensions of local user characteristics under the above occupational characteristics in the X, Y, and Z dimensions: local average call duration, local call count, local base station dwell time, local base station pass-through count, and traffic generated by related apps.

[0100] A convolutional neural network was constructed to extract features from user occupation feature maps, identifying logistics personnel with three different contact types: near, medium, and far. Company identification was then performed based on APP access type and related social circle models. The top 5 base stations with the highest number of communication interactions between the identified logistics personnel and each user on a different network were selected as candidate base stations.

[0101] A trajectory map is constructed for each user from another network and their interactions with different logistics personnel. At 60-minute intervals, the number of calls and cumulative call duration at a specific base station are recorded for seven days a week. A triple LSTM model is used to extract communication behavior characteristics between users from other networks and logistics personnel within the local network during specific time periods. The trained model is then used to discriminate the communication behavior between users from other networks and logistics personnel within the local network, identifying the base station information of users from other networks interacting with users within the local network during specific time periods. The signaling data of base stations interacting with users within the local network during specific time periods is further analyzed to more accurately identify the permanent residence location of the user from the other network.

[0102] In summary, compared with existing technologies, the cross-network user location method provided in this disclosure utilizes statistical communication signaling, as well as communication behavior data, location data, and working time characteristics of users within a specific time interval. It then filters users within the same network using these data to identify target users with different logistics occupations within that specific time interval. Next, it filters candidate base stations where target users and cross-network users interact. Finally, it inputs the cumulative communication behavior data of cross-network users and target users at these candidate base stations into a trained cross-network user location model to obtain the cross-network user's permanent residence location within different time periods. To avoid situations where the interaction circle cannot be effectively extracted, leading to inaccurate estimation of cross-network user locations, this method filters target users with different logistics occupations and uses communication behavior data of target users and cross-network users at different contact base stations (i.e., candidate base stations) within different time periods to accurately locate the cross-network user's permanent residence.

[0103] Based on the above Figure 1 and Figure 2 The specific implementation of the method shown in this embodiment provides a cross-network user positioning device, such as... Figure 8 As shown, the device includes: a statistics module 31, a first filtering module 32, a second filtering module 33, and an input module 34;

[0104] The statistics module 31 is used to collect statistics on communication signaling, as well as communication behavior data, location data, and working time characteristics of users on this network within a specific time interval;

[0105] The first filtering module 32 is used to filter the network users using the communication signaling, the communication behavior data, the location data and the working time characteristics, and to filter out target network users with different logistics occupations within the specific time interval;

[0106] The second filtering module 33 is used to filter candidate base stations where the target local network user and the external network user come into contact.

[0107] The input module 34 is used to input the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods.

[0108] In a specific application scenario, the first screening module 32 can be used to screen the network users using the communication signaling, the communication behavior data, the location data, and the working time characteristics, and screen out the network users with specific occupational characteristics within the specific time interval; construct an occupational feature map of the network users with specific occupational characteristics using the communication behavior data, location data, and working time characteristics of the network users with specific occupational characteristics; extract the occupational features from the occupational feature map using a convolutional neural network algorithm, and determine the target network users with different logistics occupations based on the occupational features.

[0109] In specific application scenarios, the second screening module 33 can be used to construct a base station relationship network; and use the base station relationship network to screen candidate base stations that the target local network user and the external network user can contact.

[0110] In a specific application scenario, the second screening module 33 can be used to determine multiple effective base stations that the target local user and the external user are in contact with; determine the base station distance between the multiple effective base stations using the base station relationship network; cluster the multiple effective base stations according to the base station distance to obtain the clustering result; and remove non-resident base stations from the clustering result to obtain the candidate base stations.

[0111] In specific application scenarios, the device also includes: a determination module 35, a construction module 36, and a training module 37;

[0112] The determining module 35 is used to determine the external network user based on the communication behavior data between the target local network user and different base stations;

[0113] Construction module 36 is used to construct the occupational characteristic trajectory map of the target local user based on the communication behavior data of the inter-network user and the target local user under the candidate base station;

[0114] Training module 37 is used to extract historical trajectory features of the occupational feature trajectory map, as well as historical cumulative communication behavior data and historical permanent residence data of the inter-network user and the target local network user in different historical time periods under the candidate base station; using the historical cumulative communication behavior data and the historical trajectory features as input features, and using the historical permanent residence data of the inter-network user in different historical time periods as training labels, iteratively training the inter-network user permanent residence model until the loss function is less than a preset threshold.

[0115] In a specific application scenario, the input module 34 can be used to calculate the weight coefficient of the probability of the target local user's activity under each candidate base station; use the attention mechanism and softmax function to perform weight processing on the weight coefficient to obtain the weight coefficient corresponding to different types of logistics occupations; concatenate the processed data of different types of logistics occupations and their corresponding weight coefficients, perform calculation through a fully connected layer, and use the softmax function to calculate whether the candidate base station is the permanent residence of the inter-network user.

[0116] It should be noted that other corresponding descriptions of the functional units involved in the inter-network user positioning device provided in this embodiment can be found in [reference]. Figure 1 and Figure 2 The corresponding descriptions of the Chinese methods will not be repeated here.

[0117] Based on the above, Figure 1 and Figure 2 Accordingly, this disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. Figure 1 and Figure 2 The method shown.

[0118] Based on this understanding, the technical solution of this disclosure 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, or portable hard drive) and includes several instructions to cause a computer device (such as a personal computer, server, or network device) to execute the methods of various implementation scenarios of this disclosure.

[0119] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 8To achieve the above objectives, this disclosure also provides an electronic device, configurable on a vehicle (e.g., an electric vehicle), in accordance with the illustrated virtual device embodiment. The device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 and Figure 2 The method shown.

[0120] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0121] Those skilled in the art will understand that the physical device structure provided in this disclosure does not constitute a limitation on the physical device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0122] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that this disclosure can be implemented using software plus necessary general-purpose hardware platforms, or it can be implemented through hardware. Compared with the prior art, the inter-network user positioning method, device, and electronic equipment provided by this disclosure utilize statistical communication signaling, as well as communication behavior data, location data, and working time characteristics of users within a specific time interval; it filters users within the network using communication signaling, communication behavior data, location data, and working time characteristics to identify target users within the network with different logistics occupations within a specific time interval; it filters candidate base stations where target users within the network contact inter-network users; and it inputs the cumulative communication behavior data of inter-network users and target users within the candidate base stations into a trained inter-network user permanent residence model to obtain the permanent residence of inter-network users in different time periods. To avoid the inability to effectively extract the social circle, which would lead to the inability to accurately estimate the permanent residence of inter-network users, this method filters target users within the network with different logistics occupations and uses the communication behavior data of target users within the network and inter-network users within different time periods at different contact base stations (i.e., candidate base stations) to accurately locate the permanent residence of inter-network users.

[0124] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus 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 apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0125] The above are merely specific embodiments of this disclosure, enabling those skilled in the art to understand or implement this disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for locating users on a different network, characterized in that, The method includes: Statistical communication signaling, as well as communication behavior data, location data, and working time characteristics of users on this network within a specific time interval; The network users are filtered using the communication signaling, communication behavior data, location data, and working time characteristics to identify target network users with different logistics occupations within the specific time interval. Filter candidate base stations where the target local network user and external network user can interact; The cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station is input into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods. The step of inputting the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods includes: Calculate the weighting coefficient of the probability of the target local user's activity under each candidate base station; By using an attention mechanism and a softmax function to process the weight coefficients, the weight coefficients corresponding to different types of logistics occupations are obtained. The data of different types of logistics occupations and their corresponding weight coefficients are concatenated, calculated through a fully connected layer, and the softmax function is used to calculate whether the candidate base station is the permanent residence of the user on the other network.

2. The method according to claim 1, characterized in that, The process of filtering network users using the communication signaling, communication behavior data, location data, and working time characteristics to identify target network users with different logistics occupations within a specific time interval includes: The network users are filtered using the communication signaling, communication behavior data, location data, and working time characteristics to identify network users with specific occupational characteristics within a specific time interval. By analyzing the communication behavior data, location data, and working time characteristics of users with specific occupational characteristics on this network, we can construct an occupational characteristic map of users with specific occupational characteristics on this network. The occupational features in the occupational feature map are extracted using a convolutional neural network algorithm. Based on these occupational features, the target users of this network with different logistics occupations are identified.

3. The method according to claim 1, characterized in that, The process of screening candidate base stations for contact between the target local network user and the external network user includes: Construct a base station relationship network; The candidate base stations that can be contacted by the target local network user and the external network user are selected using the base station relationship network.

4. The method according to claim 3, characterized in that, The step of using the base station relationship network to screen candidate base stations that allow the target local network user to contact the external network user includes: Identify multiple valid base stations through which the target user on the local network and the user on the other network come into contact; The base station distances between the plurality of valid base stations are determined using the base station relationship network; The multiple valid base stations are clustered based on the distance between them to obtain the clustering results; The candidate base stations are obtained by removing the non-resident base stations from the clustering results.

5. The method according to claim 1, characterized in that, The method further includes: The user from another network is determined based on the communication behavior data between the target user on this network and different base stations; Based on the communication behavior data of the user from the other network and the target user from the same network under the candidate base station, a career characteristic trajectory map of the target user from the same network is constructed.

6. The method according to any one of claims 1 to 5, characterized in that, The method also includes a training method for the heterogeneous user residence location model, including: Extract the historical trajectory features of the occupational feature trajectory map, as well as the historical cumulative communication behavior data and historical permanent residence data of the inter-network user and the target local network user in different historical time periods under the candidate base station; The historical cumulative communication behavior data and the historical trajectory features are used as input features, and the historical permanent residence of the inter-network user in different historical time periods is used as training labels. The inter-network user permanent residence model is iteratively trained until the loss function is less than a preset threshold.

7. A cross-network user positioning device, characterized in that, The device includes: The statistics module is used to collect statistics on communication signaling, as well as communication behavior data, location data, and working time characteristics of users on this network within a specific time interval. The first filtering module is used to filter the network users using the communication signaling, the communication behavior data, the location data, and the working time characteristics, and to filter out target network users with different logistics occupations within the specific time interval; The second filtering module is used to filter candidate base stations where the target local network user and the external network user come into contact. The input module is used to input the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods. The step of inputting the cumulative communication behavior data of the inter-network user and the target local network user under the candidate base station into the trained inter-network user permanent residence model to obtain the permanent residence of the inter-network user in different time periods includes: Calculate the weighting coefficient of the probability of the target local user's activity under each candidate base station; By using an attention mechanism and a softmax function to process the weight coefficients, the weight coefficients corresponding to different types of logistics occupations are obtained. The data of different types of logistics occupations and their corresponding weight coefficients are concatenated, calculated through a fully connected layer, and the softmax function is used to calculate whether the candidate base station is the permanent residence of the user on the other network.

8. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

Images