Different network outbound marketing behavior identification method and device, equipment and storage medium

Abstract

The invention discloses a different network outbound marketing behavior identification method, device and equipment and a storage medium, and relates to the technical field of big data, and the method comprises the steps: carrying out the preprocessing of operator system data, and obtaining the processed call information; constructing a complex social network model through a random walk model algorithm and the call information, wherein edges of the complex social network model are communication relationships among network nodes; making and recognizing the complex social network model, and determining competing nodes; and outputting different network outbound data between the home network user and the different network user based on the competition node. According to the method, the complex social network model related to the call information is constructed through the random walk model algorithm, so that the richness and the accuracy of counter recognition in the call information can be improved, and the condition that the recognition is incomplete due to the fact that an existing scheme depends on the preset number segment and the call duration is avoided; therefore, the different network outbound data between the home network user and the different network user can be identified more effectively, and more comprehensive support is provided for telecom operators.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a method, device, equipment and storage medium for identifying cross-network outbound marketing behavior. Background Art

[0002] In the telecommunications industry, cross-network outbound marketing refers to the staff of a telecommunications operator conducting marketing promotions to users of other operators through outbound calls, attracting users of other networks to switch networks or use specific services of the operator. In order to avoid customer loss, it is necessary to identify cross-network outbound marketing behaviors so as to provide precise care for users.

[0003] In existing technical solutions, the query of cross-network outbound marketing activities mainly involves preliminary screening of call records and manual verification to identify numbers that may be involved in marketing activities.

[0004] However, existing solutions mainly rely on preset number segments and call duration thresholds for preliminary screening. Although this method is simple and easy to implement, the rules are too simple and it is difficult to fully capture complex marketing and subversion behaviors. Summary of the invention

[0005] The main purpose of this application is to provide a method, device, equipment and storage medium for identifying cross-network outbound marketing behavior, aiming to solve the technical problem that existing solutions rely on preset number segments and call duration thresholds for screening, making it difficult to fully capture complex marketing and subversion behaviors.

[0006] To achieve the above purpose, the present application proposes a method for identifying cross-network outbound marketing behavior, the method comprising:

[0007] Pre-process the operator system data to obtain the processed call information between users of the same network and users of other networks;

[0008] Constructing a complex social network model by using a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes;

[0009] Conduct counter-attack identification on the complex social network model to determine competing nodes;

[0010] Based on the competing node, the inter-network outbound call data between the users of the local network and the users of the inter-network is output.

[0011] In one embodiment, the step of constructing a complex social network model using a random walk model algorithm and the call information includes:

[0012] Determining a network node of an initial network according to a call number of the call information;

[0013] Performing network division according to the distance of any adjacent community corresponding to the network node to obtain the adjacent community distance;

[0014] A complex social network model is constructed based on the network nodes and the adjacent community distances.

[0015] In one embodiment, the step of performing network division according to the distance of any adjacent community corresponding to the network node to obtain the adjacent community distance includes:

[0016] Determine the distance between each network node;

[0017] Perform network partitioning on each network node to obtain multiple communities;

[0018] Determine the distance between any adjacent communities in each community according to the distance between each network node;

[0019] Update the network partition, return to execute the step of performing network partition on each network node to obtain multiple communities, until a preset number of cycles is reached to obtain a partition result sequence;

[0020] The division result sequence is judged by a weighted modularity function to determine the adjacent community distance corresponding to the initial network.

[0021] In one embodiment, the step of identifying counter-attacks on the complex social network model and determining competing nodes includes:

[0022] Identify the defection nodes in the complex social network model according to the neighborhood information;

[0023] Perform edge detection on the candidate defection call set corresponding to the defection node to determine a candidate competing defection set;

[0024] The competing node is determined according to the distance of each node in the candidate competing counter-attack set.

[0025] In one embodiment, the step of determining a competing node according to the distance of each node in the candidate competing countermeasure set includes:

[0026] Determine the distance between each node in the candidate counter-attack set and its neighboring nodes;

[0027] Weighted average of the distances between each node and its neighboring nodes to obtain an outlier score;

[0028] According to the fault result of the outlier score, a competing node is determined.

[0029] In one embodiment, the step of preprocessing the operator system data to obtain processed call information between users of the local network and users of different networks includes:

[0030] Construct a call information input table based on operator system data;

[0031] Performing data cleaning and data transformation on the call information input table to obtain target communication data;

[0032] Extracting features of call data between users of the same network and users of other networks in the target communication data to obtain call features;

[0033] The call features are standardized to obtain processed call information.

[0034] In addition, to achieve the above purpose, the present application also proposes a device for identifying outbound marketing behavior on a different network, the device comprising:

[0035] The data processing module is used to pre-process the operator system data and obtain the processed call information between users of the same network and users of other networks;

[0036] A network construction module, used to construct a complex social network model by using a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes;

[0037] A competitor node module, used to identify counter-attacks on the complex social network model and determine competitor nodes;

[0038] The outbound call identification module is used to output the inter-network outbound call data between the users of the local network and the users of the inter-network based on the competing node.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a device for identifying cross-network outbound marketing behavior, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the method for identifying cross-network outbound marketing behavior as described above.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the method for identifying cross-network outbound marketing behavior as described above are implemented.

[0041] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the method for identifying cross-network outbound marketing behavior as described above.

[0042] One or more technical solutions proposed in this application have at least the following technical effects: this application first pre-processes the operator system data to obtain the processed call information between users of this network and users of other networks; then a complex social network model is constructed through a random walk model algorithm and the call information, and the edges of the complex social network model are the communication relationships between network nodes; then the complex social network model is subjected to counter-revolution identification to determine the competing nodes; finally, based on the competing nodes, the outbound call data between users of this network and users of other networks is output. Since this application constructs a complex social network model related to call information through a random walk model algorithm, the richness and accuracy of counter-revolution identification in call information can be improved, avoiding the situation where the existing solutions rely on preset number segments and call durations for screening and are difficult to fully identify, thereby more effectively identifying the outbound call data between users of this network and users of other networks, and providing more comprehensive support for telecom operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0045] Figure 1 A flow chart of the overall technical solution provided for this application;

[0046] Figure 2 A flowchart of the first embodiment of the method for identifying cross-network outbound marketing behavior in this application;

[0047] Figure 3 This is an example diagram of a network community provided in Example 1 of the present application;

[0048] Figure 4 A flow chart of the second embodiment of the method for identifying outbound marketing behavior on different networks provided in this application;

[0049] Figure 5 This is an example diagram of a candidate counter-call set provided in Embodiment 2 of the present application;

[0050] Figure 6 This is a schematic diagram of the module structure of the device for identifying outbound marketing behavior on a different network according to an embodiment of the present application;

[0051] Figure 7A schematic diagram of the device structure of the hardware operating environment involved in the method for identifying cross-network outbound marketing behavior in an embodiment of the present application.

[0052] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0054] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0055] Specifically, in the existing technical solutions, the query of cross-network outbound marketing behavior is mainly implemented in the following ways: First, the call records are preliminarily screened. Through the preset number segments and call duration thresholds, the numbers that may be involved in marketing behavior are identified. This method is based on simple rule matching and can preliminarily filter out suspected marketing numbers. Then comes feature extraction, which focuses on the two basic information of call duration and call frequency. Although these basic information are simple, they reflect the usage pattern of the number to a certain extent, which helps to preliminarily determine whether it involves marketing behavior. Next is data processing and analysis. This stage mainly relies on manual data cleaning and verification, including steps such as removing duplicate data, correcting erroneous data, and verifying data consistency. Finally, the identification results are presented in the form of tables and simple charts, showing basic information such as call duration and call frequency of suspected marketing numbers, so that telecom operators can intuitively understand the overview of marketing behavior.

[0056] However, due to the limitations of existing technical solutions in terms of feature extraction, data processing and analysis, and result display, the accuracy and comprehensiveness of the recognition results are greatly affected, and there are the following disadvantages:

[0057] 1. In terms of preliminary screening, it mainly relies on preset number segments and call duration thresholds for preliminary screening. Although it is simple and easy to implement, the rules are too simple and it is difficult to fully capture complex marketing and subversion behaviors.

[0058] 2. In terms of feature extraction, since it only focuses on the two basic dimensions of call duration and call frequency, its ability to capture the complex characteristics of marketing behavior is greatly limited. Because marketing behavior is often accompanied by more diversified characteristics, such as call time distribution, diversity of call objects, and even specific patterns of call content, these are ignored by existing solutions. Therefore, the accuracy and comprehensiveness of the analysis based on such a limited feature set are naturally greatly reduced, which limits the depth and accuracy of the analysis, easily misses real marketing behaviors, and may also misreport non-marketing behaviors, resulting in waste of resources and decision-making errors.

[0059] 3. In the data processing and analysis process, due to the high degree of manual dependence, the problems of low efficiency and high cost are highlighted when faced with massive amounts of data. More importantly, human factors will inevitably introduce errors, such as fatigue, negligence or misunderstanding, which may seriously affect the quality of data analysis. In addition, manual processing cannot achieve real-time or near real-time data analysis, making the solution incapable of coping with sudden or rapidly changing marketing behaviors, affecting the accuracy of the final results.

[0060] 4. In terms of result display, the existing solutions mainly adopt the form of tables and simple charts. Although this display method is intuitive, it lacks depth and insight. It cannot fully display the internal connection between data, trend changes or comparative analysis, making it difficult for telecom operators to fully understand the nature and laws of marketing behavior from multiple dimensions. This limitation restricts telecom operators from accurately judging and effectively responding to marketing behavior, making it difficult to directly support business decisions, and affecting the formulation and implementation of their market strategies.

[0061] In summary, the limitations of existing technical solutions in feature extraction, data processing and analysis, and result display together constitute their significant shortcomings. These shortcomings not only affect the accuracy and comprehensiveness of the recognition results, but also limit the effectiveness and practicality of the solutions, making telecom operators face many challenges when dealing with cross-network outbound marketing behaviors.

[0062] To solve the above problems, this application proposes a method for identifying cross-network outbound marketing behavior to improve the query and monitoring capabilities of cross-network outbound marketing behavior. Figure 1 As shown, Figure 1 Flow chart of the overall technical solution provided for this application. The technical solution of this application is divided into two parts:

[0063] The first part is model building, which uses historical data transfer as positive samples and builds a complex social network model based on cross-network call data to identify high-risk calls and peer numbers that are suspected of being subversion calls.

[0064] The second part is the solution output, which outputs the competing numbers of different networks and the user numbers of Yicefang's own network to the label platform, and the business party can use it by accessing the label library.

[0065] This application can improve the richness and accuracy of feature extraction, automate data processing and analysis, and deepen the display and application of results by building a complex social network model, more effectively identify competing numbers on other networks and numbers of users who are easy to defect to the network, and provide telecom operators with more comprehensive and in-depth insights and support. This application will be described in detail below.

[0066] It should be noted that the execution subject of this embodiment can be a computing service device with data preprocessing, social network model construction and cross-network outbound marketing behavior recognition functions, such as a personal computer, server, etc., or an electronic device that can realize the above functions, a cross-network outbound marketing behavior recognition device (referred to as recognition device) that executes the cross-network outbound marketing behavior recognition method of this application, etc., and this embodiment does not limit this. The following takes the recognition device as an example to illustrate this embodiment and the following embodiments.

[0067] Based on this, the embodiment of the present application provides a method for identifying outbound marketing behavior on a different network, referring to Figure 2 , Figure 2 A flow chart of the first embodiment of the method for identifying cross-network outbound marketing behavior in this application.

[0068] In this embodiment, the method for identifying outbound marketing behavior on a different network includes steps S10 to S40:

[0069] Step S10: pre-processing the operator system data to obtain processed call information between users of the same network and users of other networks.

[0070] It should be noted that operator system data refers to data generated and collected by communication operators during their operations, such as user call data, portability data, number ownership data, portability numbers and normal numbers, and the current portability status of numbers.

[0071] It should be noted that call information is information between local network users and users of other networks obtained after pre-processing the operator system data (such as data clarification, sample screening, and data processing). Local network users are users who belong to a specific communication network (such as the network of a certain telecom operator). Users of other networks are users who do not belong to this specific network but belong to other communication networks (networks of other telecom operators).

[0072] By using the existing call data, porting data, number ownership data and other information in the operator system, the cross-network call records of the ported number one month before the porting are used as positive samples to build the complex social network model described below, which can identify high-risk calls and peer numbers suspected of subversion calls.

[0073] Specifically, Figure 1 As shown in the figure, this step is the process of data preparation and processing, which is used to select indicators and standardize data for the operator system data. In the preprocessing process, the useless data in the operator system data can be cleaned first, and then the call data related to cross-network marketing can be selected, and finally the call data related to cross-network marketing can be feature extracted to obtain the processed call information between users of this network and users of cross-networks.

[0074] Furthermore, during the data preprocessing process, manual intervention steps may be added, such as manual review, parameter adjustment, etc., to further improve the accuracy of data processing, which is not limited in this embodiment.

[0075] In this implementation, by preprocessing the operator system data, call information between users of the same network and users of other networks can be obtained to improve the accuracy and depth of analysis.

[0076] In a feasible implementation, step S10 of this embodiment may include the steps of: constructing a call information input table based on operator system data; performing data cleaning and data transformation on the call information input table to obtain target communication data; performing feature extraction on the call data between users of the same network and users of other networks in the target communication data to obtain call features; and performing standardization processing on the call features to obtain processed call information.

[0077] Specifically, the process of data preparation and preprocessing may include the following steps:

[0078] First, data input. Based on the information in the operator system data, a call information input table can be constructed to form the input data required for this embodiment. The call information input table includes user ID, peer number ID, call duration, call time, and peer number ownership. The call information input table includes call data of the ported number and the normal number, and the call data may include the current ported status of the number.

[0079] Then, data cleaning is performed. For example, the call information input table may contain noise data caused by machine failure, such as null values ​​and records of calls that are too long, which should be deleted.

[0080] Then, data transformation is performed. For example, when a government or enterprise customer dials an external number, the called number starts with 9, and it needs to be restored to the original number. Through the above data cleaning and data transformation, the target communication data can be obtained.

[0081] The next step is to select samples. This embodiment is to identify the outbound marketing behavior of operators' customers being poached by competitors, so the selected samples are the call data between customers of the same network and customers of other networks.

[0082] The next step is feature extraction. In addition to call duration and frequency, more dimensional features can be considered, such as call time distribution, the relationship between call objects, etc., to improve the accuracy and depth of analysis. Specifically, after selecting call data between customers on the same network and customers on other networks, all call records between the two customers in the month are summarized to construct call features such as the number of calls, call duration, average call duration, number of calling calls, calling call duration, number of called calls, and called call duration, to improve the richness and accuracy of feature extraction.

[0083] Finally, feature processing is performed. Since the call duration data has large fluctuations in value, in order to ensure that the distribution differences between features are roughly the same, the call duration can be standardized to obtain processed call information. This can avoid excessive influence of a single variable and facilitate model convergence.

[0084] The above process can be carried out by automated means, such as automated data cleaning, deduplication, verification, etc., to reduce reliance on manual work, thereby significantly improving data processing efficiency. At the same time, automated processing also reduces errors caused by human factors and improves data accuracy and reliability.

[0085] Furthermore, in the feature extraction stage, the key features mentioned above (such as call time distribution and call object relationship) can be avoided, and other non-key features with lower correlation can be selected. The extracted non-key features are then fused, transformed or weighted so that the new feature set is significantly different from the processed call information in form as a data substitute.

[0086] In this embodiment, when processing the call information in the operator system data, in addition to the traditional call duration and frequency, multi-dimensional feature extraction is also performed, including call duration, average call duration, number of calling calls, calling call duration, number of called calls, called call duration, etc. These features can more carefully describe the user's call behavior, enhance the depth and accuracy of the analysis, and make the recognition results more accurate and comprehensive.

[0087] Step S20: constructing a complex social network model through a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes.

[0088] It should be noted that the random walk model algorithm is an algorithm used to describe uncertain and random processes such as call information.

[0089] Because the complex social network model to be constructed in this embodiment is a call information network, the network nodes represent the numbers at both ends of the call, and the edges represent the correlation between the numbers (i.e., the communication relationship). For the selection of the model algorithm, the characteristics of the operator data are taken into consideration: ① The operator industry is a data-intensive enterprise, and the algorithm should be able to process large-scale data; ② The closeness of customer calls is different, and the algorithm should be applicable to weighted networks; ③ Based on the actual social structure, customers will form groups of different sizes, and the algorithm should be able to discover communities of different sizes.

[0090] Therefore, according to the above characteristics, the appropriate algorithm to be selected is the random walk model algorithm, which is used as a network partitioning method. The time complexity of this algorithm for sparse networks is O(m2 logn) (where n and m are the number of network nodes and edges respectively), which can handle large-scale networks with millions of nodes and is suitable for weighted networks and discovering communities of different sizes.

[0091] Specifically, Figure 1 As shown, this step is the stage of building an inter-network call network, which is mainly built through the above call data. In addition to building the above-mentioned complex social network model, other types of network models different from the complex social network model can also be built, such as simple graph models, weighted graph models, or models based on other relationships (such as text messages, social media interactions). This can further reduce the complexity of the model by simplifying the model structure or reducing the number of model parameters.

[0092] In this embodiment, by constructing a complex social network model, the call relationships between users can be captured more comprehensively, and suspected subversion calls and high-risk calls hidden in the complex network can be identified.

[0093] In a feasible implementation, step S20 of this embodiment may include the steps of: determining the network nodes of the initial network according to the call number of the call information; dividing the network according to the distance of any adjacent community corresponding to the network node to obtain the adjacent community distance; and constructing a complex social network model based on the network node and the adjacent community distance.

[0094] Specifically, the steps for building a complex social network model based on call information are as follows:

[0095] First, all network nodes may be determined according to the call number of the call information obtained after processing in step S10.

[0096] Then, all network nodes can be divided into communities (i.e., network partitioning). First, the community can be initialized so that each network node is a community (a community is a group of closely connected network nodes composed of multiple network nodes), and the distances between different communities are calculated; then, any adjacent communities are gradually merged, and the distances between different communities after the merger are calculated again (i.e., the distance between adjacent communities and communities).

[0097] Finally, a complex social network model is constructed based on all the above network nodes and the calculated distances.

[0098] In this embodiment, by constructing a complex social network model including multiple nodes (users) and multiple sides (call relationships), it is possible to more comprehensively capture the call behavior patterns and potential relationships between users, thereby more accurately identifying suspected subversion calls and high-risk calls.

[0099] In a feasible implementation, the step of performing network division according to the distance of any adjacent community corresponding to the network node in this embodiment to obtain the distance of adjacent communities includes: determining the distance between each network node; performing network division on each network node to obtain multiple communities; determining the distance of any adjacent community in each community according to the distance between each network node; updating the network division, returning to execute the step of performing network division on each network node to obtain multiple communities until a preset number of cycles is reached to obtain a division result sequence; judging the division result sequence by a weighted modularity function to determine the distance of adjacent communities corresponding to the initial network.

[0100] It should be noted that the preset number of cycles is the preset number of community divisions. By dividing the communities multiple times, the network division with the best effect can be selected as the complex social network model.

[0101] Specifically, the steps for calculating the distance between network nodes are as follows:

[0102] (1) Calculate the distance between network nodes i and j as:

[0103]

[0104] Where d(k) is the degree of node k (i.e., the number of edges directly connected to the node), is the probability of going from node i to node k in t steps, is the probability of going from node j to node k in t steps.

[0105] (2) Each network node is divided into multiple communities (the community structure refers to the grouping of closely related nodes in the network. The nodes in these groups are relatively tightly connected, while the connections between groups are relatively sparse). The probability of going from community C to node j is defined as the average probability of any node i to j in C, that is:

[0106]

[0107] Among them, |C| is the number of nodes in community C, is the probability of going from node i to node j in t steps.

[0108] (3) If any communities C1 and C2 have at least one edge connecting them, calculate the distance between communities C1 and C2:

[0109]

[0110] in, (where i = 1, 2, ...) is the probability of going from community C to node j obtained in the previous step, and d(k) is the degree of node k.

[0111] (4) Treat each node as a separate community, and divide the initial network into P1 = {{v}, v∈V} (v is all the nodes in the grid, and P1 is the set of all the nodes in the network). The distance of any adjacent community is calculated according to formula (3).

[0112] (5) For the h-th partition result P h ,according to:

[0113]

[0114] Among them, Δσ is a representative of the change, which is used here to refer to the distance between two communities; the standard selects the two communities C1 and C2 that make Δσ the smallest to merge, where is the square of the distance between adjacent communities C1 and C2 obtained in the previous step; n is the number of cycles = h (i.e. the preset number of cycles), and |C| is the number of nodes in community C.

[0115] (6) Update the network partition into P h+1 =(P h \{C1,C2})∪C3, calculate the adjacent community distance.

[0116] (7) Repeat steps (5) and (6) n-1 times to form a network partition result sequence (P h ) i≤h≤n , using the weighted modularity function (weighted modularity is an indicator to measure the quality of the current network partitioning, through which the best network partitioning effect can be evaluated) to determine (P h ) i≤h≤n The best network partition is obtained, and the distance of each community in the best network partition is obtained. In order to find the best network partition, it is necessary to traverse all possible partition combinations and calculate their weighted modularity. In each iteration, two communities are merged to maximize the increase in weighted modularity. This process is repeated until the predetermined number of communities n is reached.

[0117] For ease of understanding, the best network partitioning effect is evaluated by weighted modularity function as follows Figure 3 As shown, Figure 3 This is an example diagram of a network community provided in Example 1 of the present application. Based on the nodes and distances obtained above, taking the node sequence 1 to 115 as an example, the complex social network model with the best network division effect that can be constructed is as follows:

[0118] (1) As shown in the figure, nodes of the same color are grouped together, and the internal nodes are closely connected communities. The nodes of the light green community may include the sequence: 109, 9, 79, 112, 22, 23, 69, 52, 8, 78; the nodes of the light blue community include the sequence: 84, 68, 74, 47, 111, 89, 50, 54, 115; the nodes of the red community may include the sequence: 91, 41, 51, 82, 98, 85, 29, 75, 4, 11, 99, 12, 6, 103, 73, 70, 108, 53; the nodes of the dark blue community include the sequence: 48, 65, 33, 14, 101, 16, 3, 40, 7, 107, 61; the nodes of the pink community include the sequence: 17, 105, 94, 5, 10, 24, 1 , 42; the nodes of the dark yellow community include the sequence: 93, 92, 58, 87, 49, 113, 45, 67, 76; the nodes of the yellow community include the sequence: 55, 83, 30, 72, 102, 43, 32, 80, 81, 36, 100, 56, 31, 36, 35, 20, 95, 62, 19; the nodes of the dark green community include the sequence: 38, 26, 104, 110, 2, 46, 34, 106, 90; the nodes of the brown community include the sequence: 64, 21, 96, 86, 27, 13, 18, 57, 39, 114, 60, 66, 77, 37, 63, 71, 88, 28, 44, 97.

[0119] (2) There is a single point that is connected to other communities but not closely, such as nodes 25, 59, or 15.

[0120] Through the above community detection, the community structure in the network can be discovered, which helps to understand the relationship in the network, discover the pattern of the network, identify key nodes and clusters, etc.

[0121] In this implementation, the process of building a social network model based on the random walk model algorithm combined with operator information includes determining network nodes based on call information of numbers; calculating the distance between nodes, the distance between nodes and communities, and the distance between communities; continuously updating the network community divisions and determining the best network community divisions; building a social network model based on nodes and distances, and the edges of the social network model are the call relationships between nodes. Complex social network models can achieve comprehensive capture and in-depth analysis of user call relationships, which is the key to improving the accuracy of identifying high-risk calls and subversion behaviors.

[0122] Further, such as Figure 1 As shown in the figure, after constructing a complex social network model, the authenticity of the network can also be tested based on the static statistical characteristics and topological properties of the network. The network authenticity test is specifically as follows:

[0123] The weighted modularity function mentioned above is used as the static statistical feature of the network to evaluate the formed communities. The modularity measures whether the nodes in the community have as many edges as possible and the edges between communities are as few as possible. For weighted networks, the weighted modularity formula is used:

[0124]

[0125] Where W is the sum of network edge weights; A ij is the network adjacency matrix; S i and S j are the edge weights and the weights of nodes i and j respectively; C i and C j are the communities where nodes i and j are located respectively. When C i =C j When δ(C j , C j )=1, otherwise it is 0. The Q value is between [0,1], and Q=0.3 is the lower limit of the network with obvious community structure, at which point the network is real.

[0126] Step S30: performing counter-attack identification on the complex social network model to determine competing nodes.

[0127] It should be noted that the competing node is a node in the complex social network model that is abnormally related to the outbound calling behavior of different networks compared with other nodes. Among them, the competing node can include two possibilities, one is the user of the local network, and the other is the user of the different network, in which the user of the different network is the competing number.

[0128] Specifically, the process of counter-attack identification may include: first, the distance between each node and its neighboring nodes may be determined. Then, the distance between the neighbors of each node may be measured using anomaly score values. Then, based on the distribution of anomaly score values, nodes with obvious anomalies may be determined and used as competing nodes.

[0129] Through the competitive node, it is possible to determine the users on the other network who are involved in the outbound marketing activities and their competitive numbers, as well as the user nodes on this network connected to the nodes corresponding to the competitive numbers (i.e., users who are easily defected).

[0130] In this implementation, countermeasure counter-identification can be used to quickly and accurately identify users on other networks who are related to outbound marketing behaviors in a complex social network model.

[0131] Step S40: Based on the competing node, outputting the inter-network outbound call data between the users of the local network and the users of the inter-network.

[0132] It should be noted that the inter-network outbound call data is data related to outbound marketing activities between users of this network and users of other networks, such as competing numbers, operators, locations, the counterpart numbers of recent calls with users of this network, number of calls, information on numbers of users of this network that have been poached, number of outbound calls, etc. This embodiment does not impose any restrictions on this.

[0133] In this implementation, by identifying competing nodes, the model result table is updated and output on a daily basis. The model result table outputs competing numbers, competing number anomalies, and high-risk call information such as numbers poached from this network. It not only contains basic information, but can also perform in-depth analysis based on actual business conditions, providing telecom operators with more comprehensive and in-depth insights to support their business decision-making and strategy formulation.

[0134] In the technical solution provided in this embodiment, the useless data in the operator system data can be cleaned first, and then the call data related to cross-network marketing can be selected. Finally, the call data related to cross-network marketing can be feature extracted to obtain the processed call information between users of this network and users of other networks. Then, a complex social network model is constructed through the random walk model algorithm and call information. The edges of the complex social network model are the communication relationships between network nodes, so as to more comprehensively capture the call behavior patterns and potential relationships between users. Then, the anomaly score value can be used to measure the distance between the neighbors of each node, and the nodes with obvious anomalies can be determined according to the distribution of the anomaly score value, and used as competing nodes. Finally, by identifying the competing nodes, the model result table is updated and output on a daily basis. The model result table outputs high-risk call information such as competing numbers, competing number anomalies, and poaching of numbers on this network. Since this embodiment constructs a complex social network model related to call information through a random walk model algorithm, it can improve the richness and accuracy of subversion identification in call information, avoids the situation where existing solutions rely on preset number segments and call duration for screening and are difficult to fully identify, and can more effectively identify inter-network outbound call data between users of this network and users of other networks, providing more comprehensive support for telecom operators.

[0135] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be described in detail later. Figure 1 As shown, this embodiment is a competitive counter-attack call identification process, which mainly divides the community structure, identifies counter-attack nodes, and generates a list of high-risk calls and cross-network competitive numbers, as described in detail below.

[0136] On this basis, please refer to Figure 2 and Figure 4 , Figure 4 A flow chart of the second embodiment of the method for identifying cross-network outbound marketing behavior in this application.

[0137] In this example, step S30 includes steps S31 to S33:

[0138] Step S31: Identify defective nodes in the complex social network model based on neighborhood information.

[0139] It should be noted that the neighborhood information is the information related to each node.

[0140] For example, if the neighbors of node t belong to multiple different communities, then node t is likely to be a defection node. Among them, nodes that may be defections have the following characteristics: nodes that have calls with multiple unrelated groups have a higher probability of being defection calls (that is, nodes that have edges with multiple communities have a higher probability of being defection nodes).

[0141] Step S32: performing edge detection on the candidate defection call set corresponding to the defection node to determine a candidate competing defection set.

[0142] It should be noted that the candidate defection call set is the set of the above-mentioned defection nodes.

[0143] It should be noted that the candidate competitive counter-attack set is a further screening of the candidate counter-attack call set to obtain a node set related to the outbound calling behavior.

[0144] For ease of understanding, for example Figure 5 As shown, Figure 5 This is an example diagram of a candidate defection call set provided in Example 2 of the present application. First, a node set q that is not in the same community but has an edge can be found, and this node set q is used as a candidate defection call set.

[0145] The node sets divided by the two circles in the figure represent two different communities (one circle has node numbers: 12, 10, 5, 9, 11, 8, 6, 7, and the other circle has node numbers: 3, 1, 4, 2). As can be seen from the figure, although they are in different communities, there are edges between the two communities. There is an edge between node 6 and node 4, and there is an edge between node 5 and node 1. Therefore, nodes 1, 4, 5, and 6 are considered bridge nodes, and the node set they form is the candidate counter-strategy set.

[0146] Step S33: Determine the competing node according to the distance of each node in the candidate competing counter-attack set.

[0147] Specifically, after determining the candidate counter-attack set, the nodes with obvious abnormalities can be determined according to the distribution of the distance between each node and its neighboring nodes, and they can be used as counter-attack nodes.

[0148] In this implementation, by determining a candidate competitive countermeasure set, competitive nodes related to cross-network outbound marketing behaviors can be clearly discovered.

[0149] In a feasible implementation, step S33 of this example includes the steps of: determining the distance between each node in the candidate competitive counter-strategy set and its neighboring nodes; weighted averaging the distance between each node and its neighboring nodes to obtain an outlier score; and determining the competitive node based on the fault result of the outlier score.

[0150] It should be noted that the outlier score is a value that measures the distance between each node and its neighbor nodes, which can be expressed as a weighted average of similarities.

[0151] Specifically, the distance between neighbor nodes of each candidate node in the candidate competitive counter-attack set can be calculated according to the formula for calculating the distance of the node in the above step S20. The larger the distance between the neighbor nodes of the candidate node, the smaller the connectivity between the neighbor nodes, and the greater the probability that the candidate node is a competitive node.

[0152] The outlier score can then be used to measure the distance between the neighbors of each candidate node. Since the number of neighbors of a candidate node must be greater than or equal to 2, the outlier score can be calculated by taking the weighted average of the similarities of each neighbor node pair (u, v)∈E(i) of candidate node i, as follows:

[0153]

[0154] Among them, Q represents the modularity of node i, E(i) represents the neighbor nodes of node i, and S represents the distance of the node. The larger E is, the higher the possibility that it is a competing node. Finally, the abnormal score values ​​of the candidate nodes are ranked from small to large, and the abnormal score values ​​E of the candidate nodes ranked in the top N (the threshold can be adjusted according to the results) are selected for analysis and distribution. According to the fault results of E, the distance between the nodes with faults and their neighbor nodes is large, and compared with other nodes, they are obviously abnormal. These nodes are divided into competing nodes. Among them, competing nodes include two possibilities, one is the user of this network, and the other is the user of another network, where the user of another network is a competing number. The user node of this network connected to the competing number node (with call records) is an easy-to-rebel user, and other user nodes of another network connected to the easy-to-rebel user node are potential competing nodes (which need to be paid special attention to in business).

[0155] In the technical solution provided in this embodiment, the process of determining competing nodes based on the constructed complex social network model includes using neighborhood information to identify defective nodes, determining competing nodes from candidate nodes based on outlier scores, and determining NetEase defective users based on the competing nodes, thereby ensuring the recognition accuracy of competing nodes.

[0156] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above introduction and will not be repeated later.

[0157] like Figure 1 As shown, after the model outputs the results, the competing numbers of different networks and the numbers of the local network can be output. At this time, business applications can be performed based on the system's competing customer service numbers, competing number anomalies, numbers of the local network, and the number of outbound calls. The specific applications are as follows:

[0158] 1. Competitive number marking and statistics: Determine the competitive number, mark the competitive customer service number through calculation, and classify it into major categories based on the abnormal values ​​of the number, such as high-risk competitive and general competitive types, and support statistical summary by province, city, district, and grid.

[0159] 2. Yicefan customer statistics: Statistics are collected based on the outbound call numbers, and the province, city, district, and grid are analyzed to conduct statistics on poached customers, understand the poaching situation in the region, and conduct separate classification and statistics on Yicefan users.

[0160] 3. Yicefan customer care: Create care tickets for Yicefan customers, carry out precise care, and avoid customer churn.

[0161] Furthermore, the results of business applications can be presented in the form of tables, charts, etc., or in the form of interactive interfaces, visual maps, etc. The analysis results can be combined with different business scenarios to provide customized solutions or suggestions.

[0162] In this implementation, not only a basic information list is provided, but also in-depth analysis is performed in combination with actual business, such as community structure division, identification of defection nodes, etc., to provide operators with comprehensive and in-depth insights. This helps telecom operators better understand market dynamics and make effective business decisions and strategies.

[0163] Furthermore, in order to reflect the performance of the method for identifying cross-network outbound marketing behavior in this application, its performance test is as follows:

[0164] 1. Performance testing plan.

[0165] (1) Test purpose: To verify the feasibility and superiority of this solution in terms of performance and effect, the algorithm running time and algorithm output list are tested.

[0166] (2) Test data selection: Four grids of different sizes are selected for testing. The test data table is shown in Table 1:

[0167] Table 1

[0168]

[0169]

[0170] (3) Test script selection: Test the entire process of the complete solution from data reading to result list output.

[0171] (4) Setting the number of test rounds: Set to 5 rounds of testing for each data, use dedicated equipment for special testing, and disconnect the network and shut down all other programs to ensure that the test environment is free of external interference.

[0172] 2. Performance test environment setting: The development environment may use Python 3.7.16, which is not limited in this embodiment.

[0173] 3. Performance test results. The test results are shown in the table:

[0174] Table 2

[0175]

[0176] According to the above results, compared with the output time of several days or even months required for the analysis of the competitive results of the prior art solutions, the performance of the solution of the present application is measured in milliseconds, and can basically achieve quasi-real-time output. And as the number of nodes (number of peer numbers) increases, the number of calls grows exponentially, but the proportion of the increase in running time has not increased significantly, which shows the superiority of the solution in terms of large data volume performance. Moreover, the recognition effect of large data volumes is excellent. As the amount of data increases, the recognition effect also increases significantly. The recognition ratio increases from 0.067% to 0.21%, which is three times the original. This solution does not require manual positive sample labeling, saving a lot of labor costs. Compared with traditional clustering algorithms, this solution is more interpretable, the threshold division does not require manual adjustment, and the portability and reproducibility are stronger.

[0177] In the technical solution provided in this embodiment, this application has higher recognition accuracy and comprehensiveness than existing solutions by constructing complex social network models, extracting multi-dimensional features, realizing automated data processing and analysis, and deepening the display and application of results. It can improve data processing efficiency and reduce human errors. Therefore, by monitoring and identifying outbound marketing behaviors of different networks, it can help users reduce the number of harassing calls they receive and improve their communication experience and satisfaction. It can also quickly understand the market dynamics and marketing strategies of competitors, so as to formulate more effective market response strategies, allocate resources and budgets more specifically, avoid unnecessary waste, and improve resource utilization efficiency.

[0178] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the method for identifying cross-network outbound marketing behavior of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0179] This application also provides a device for identifying outbound marketing behavior on a different network. Please refer to Figure 6 , Figure 6 This is a schematic diagram of the module structure of the device for identifying outbound marketing behavior on a different network according to an embodiment of the present application; the device for identifying outbound marketing behavior on a different network includes:

[0180] The data processing module 601 is used to pre-process the operator system data to obtain the processed call information between users of the same network and users of other networks;

[0181] A network construction module 602, configured to construct a complex social network model using a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes;

[0182] A competitor node module 603 is used to identify counterattacks on the complex social network model and determine competitor nodes;

[0183] The outbound call identification module 604 is used to output the outbound call data between the users of the local network and the users of the other network based on the competing node.

[0184] In this embodiment, the network construction module 602 is also used to determine the network nodes of the initial network according to the call number of the call information; divide the network according to the distance of any adjacent community corresponding to the network node to obtain the adjacent community distance; and construct a complex social network model based on the network node and the adjacent community distance.

[0185] Furthermore, the network construction module 602 is also used to determine the distance between each network node; perform network division on each network node to obtain multiple communities; determine the distance of any adjacent community in each community based on the distance between each network node; update the network division, return to execute the step of performing network division on each network node to obtain multiple communities, until a preset number of cycles is reached to obtain a division result sequence; judge the division result sequence by a weighted modularity function to determine the adjacent community distance corresponding to the initial network.

[0186] Furthermore, the competitive node module 603 is also used to identify the defection node in the complex social network model based on the neighborhood information; perform edge detection on the candidate defection call set corresponding to the defection node to determine the candidate competitive defection set; and determine the competitive node based on the distance of each node in the candidate competitive defection set.

[0187] Furthermore, the competitive node module 603 is also used to determine the distance between each node in the candidate competitive counter-strategy set and its neighboring nodes; weighted average the distance between each node and its neighboring nodes to obtain an outlier score; and determine the competitive node based on the fault result of the outlier score.

[0188] Furthermore, the data processing module 601 is also used to construct a call information input table based on the operator system data; perform data cleaning and data transformation on the call information input table to obtain target communication data; perform feature extraction on the call data between users of the same network and users of other networks in the target communication data to obtain call features; and perform standardization on the call features to obtain processed call information.

[0189] Other embodiments or specific implementation methods of the device for identifying cross-network outbound marketing behavior of the present application can refer to the above-mentioned method embodiments and will not be repeated here.

[0190] The device for identifying cross-network outbound marketing behavior provided by the present application adopts the method for identifying cross-network outbound marketing behavior in the above-mentioned embodiment, which can solve the technical problem that the existing solution relies on preset number segments and call duration thresholds for screening, and is difficult to fully capture complex marketing and subversion behaviors. Compared with the prior art, the beneficial effects of the device for identifying cross-network outbound marketing behavior provided by the present application are the same as the beneficial effects of the method for identifying cross-network outbound marketing behavior provided by the above-mentioned embodiment, and the other technical features in the device for identifying cross-network outbound marketing behavior are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0191] The present application provides a device for identifying cross-network outbound marketing behavior, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the cross-network outbound marketing behavior identification method in the above-mentioned embodiment one.

[0192] Reference below Figure 7 , Figure 7The schematic diagram of the device structure of the hardware operating environment involved in the method for identifying cross-network outbound marketing behavior in the embodiment of the present application shows a schematic diagram of the structure of a cross-network outbound marketing behavior identification device suitable for implementing the embodiment of the present application. The cross-network outbound marketing behavior identification device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The cross-network outbound marketing behavior identification device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0193] like Figure 7 As shown, the cross-network outbound marketing behavior recognition device may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the cross-network outbound marketing behavior recognition device are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the outbound marketing behavior recognition device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows an outbound marketing behavior recognition device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0194] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0195] The cross-network outbound marketing behavior identification device provided by the present application adopts the cross-network outbound marketing behavior identification method in the above-mentioned embodiment, which can solve the technical problem that the existing solution relies on preset number segments and call duration thresholds for screening, and it is difficult to fully capture complex marketing and subversion behaviors. Compared with the prior art, the beneficial effects of the cross-network outbound marketing behavior identification device provided by the present application are the same as the beneficial effects of the cross-network outbound marketing behavior identification method provided by the above-mentioned embodiment, and the other technical features in the cross-network outbound marketing behavior identification device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0196] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0197] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0198] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the method for identifying cross-network outbound marketing behavior in the above-mentioned embodiment.

[0199] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0200] The above-mentioned computer-readable storage medium may be included in the device for identifying the outbound marketing behavior of a different network; or it may exist independently without being assembled into the device for identifying the outbound marketing behavior of a different network.

[0201] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the cross-network outbound call marketing behavior identification device, the cross-network outbound call marketing behavior identification device: pre-processes the operator system data to obtain the processed call information between users of this network and users of other networks; constructs a complex social network model through a random walk model algorithm and the call information, and the edges of the complex social network model are the communication relationships between network nodes; performs counter-invention identification on the complex social network model to determine competing nodes; and based on the competing nodes, outputs the cross-network outbound call data between users of this network and users of other networks.

[0202] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0203] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0204] The modules involved in the embodiments described in the present application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0205] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned method for identifying cross-network outbound marketing behaviors, and can solve the technical problem that the existing solutions rely on preset number segments and call duration thresholds for screening, and it is difficult to fully capture complex marketing and subversion behaviors. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the method for identifying cross-network outbound marketing behaviors provided by the above-mentioned embodiments, and will not be repeated here.

[0206] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for identifying cross-network outbound marketing behavior.

[0207] The computer program product provided by this application can solve the technical problem that the existing solutions rely on preset number segments and call duration thresholds for screening, which makes it difficult to fully capture complex marketing and subversion behaviors. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the method for identifying cross-network outbound marketing behaviors provided by the above-mentioned embodiment, and will not be repeated here.

[0208] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for identifying outbound marketing behavior of a different network, characterized in that: The method includes: Pre-process the operator system data to obtain the processed call information between users of the same network and users of other networks; Constructing a complex social network model by using a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes; Conduct counter-attack identification on the complex social network model to determine competing nodes; Based on the competing node, the inter-network outbound call data between the users of the local network and the users of the inter-network is output.

2. The method according to claim 1, characterized in that The step of constructing a complex social network model using a random walk model algorithm and the call information includes: Determining a network node of an initial network according to a call number of the call information; Performing network division according to the distance of any adjacent community corresponding to the network node to obtain the adjacent community distance; A complex social network model is constructed based on the network nodes and the adjacent community distances.

3. The method according to claim 2, characterized in that The step of dividing the network according to the distance of any adjacent community corresponding to the network node to obtain the distance of the adjacent community includes: Determine the distance between each network node; Perform network partitioning on each network node to obtain multiple communities; Determine the distance between any adjacent communities in each community according to the distance between each network node; Update the network partition, return to execute the step of performing network partition on each network node to obtain multiple communities, until a preset number of cycles is reached to obtain a partition result sequence; The division result sequence is judged by a weighted modularity function to determine the adjacent community distance corresponding to the initial network.

4. The method according to claim 1, characterized in that The step of identifying counter-attacks on the complex social network model and determining competing nodes includes: Identify the defection nodes in the complex social network model according to the neighborhood information; Perform edge detection on the candidate defection call set corresponding to the defection node to determine a candidate competing defection set; The competing node is determined according to the distance of each node in the candidate competing counter-strategy set.

5. The method according to claim 4, characterized in that The step of determining a competing node according to the distance of each node in the candidate competing counter-attack set includes: Determine the distance between each node in the candidate counter-attack set and its neighboring nodes; Weighted average of the distances between each node and its neighboring nodes to obtain an outlier score; According to the fault result of the outlier score, a competing node is determined.

6. The method according to any one of claims 1 to 5, characterized in that The step of preprocessing the operator system data to obtain processed call information between users of the local network and users of other networks includes: Construct a call information input table based on operator system data; Performing data cleaning and data transformation on the call information input table to obtain target communication data; Extracting features of call data between users of the same network and users of other networks in the target communication data to obtain call features; The call features are standardized to obtain processed call information.

7. A device for identifying outbound marketing behavior on a different network, characterized in that: The device comprises: The data processing module is used to pre-process the operator system data and obtain the processed call information between users of the same network and users of other networks; A network construction module, used to construct a complex social network model by using a random walk model algorithm and the call information, wherein the edges of the complex social network model are communication relationships between network nodes; A competitor node module, used to identify counter-attacks on the complex social network model and determine competitor nodes; The outbound call identification module is used to output the inter-network outbound call data between the users of the local network and the users of the inter-network based on the competing node.

8. A device for identifying outbound marketing behavior on a different network, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the method for identifying cross-network outbound marketing behavior as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the method for identifying cross-network outbound marketing behavior as described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the method for identifying cross-network outbound marketing behavior as described in any one of claims 1 to 6 are implemented.

Images