Government and enterprise customer cross marketing method and system based on graph analysis

Abstract

The invention discloses a graph analysis-based government and enterprise customer cross-marketing method and system, and belongs to the technical field of computers. The method comprises the following steps: acquiring basic data; constructing a knowledge graph according to the basic data; performing community division on the knowledge graph to obtain a plurality of communities; based on average or similarity calculation, screening high-quality communities from all communities; and recommending the high-quality community. Through cross marketing, an enterprise can provide more product and service combinations for government and enterprise customers, and the diversified requirements of the government and enterprise customers are met. For example, while an informatization system is provided for government customers, related data security services, cloud computing resources and the like can be provided, so that the customers are more dependent on an overall solution of an enterprise, and the stickiness and loyalty of the customers are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly relates to a method and system for cross-marketing of government and enterprise customers based on graph analysis. Background Art

[0002] The continuous development of economy and technology has made enterprises with IT, CT, and DT products pay more and more attention to the rapid iteration of products and services, in order to achieve the goal of building a market moat. At the same time, due to factors such as the customer group differences and product demand differences between TOB and TOC, it poses obvious challenges to the marketing capabilities of enterprises, specifically manifested in:

[0003] 1. High decision-making complexity: The procurement decisions of government and enterprise customers usually involve multiple departments and stakeholders, and the decision-making process is complex and lengthy. For example, the procurement of government projects requires strict approval procedures, and enterprise procurement needs to consider factors such as budget and technical compatibility.

[0004] 2. Strong demand customization: The demands of different government and enterprise customers vary greatly, and often require highly customized solutions. For example, government informatization projects may need to meet specific security and functional requirements, and enterprise informatization systems need to be closely integrated with existing business processes.

[0005] 3. Fierce competition: There are many competitors in the market, and the competition strategies are diverse. Enterprises need to compete fiercely with competitors in terms of technology, price, service, etc. For example, in the IT field, many domestic and foreign manufacturers are committed to providing informatization products and services for government and enterprise customers, and the competition is very fierce.

[0006] 4. Policy risks: Changes in government policies may have a significant impact on the procurement demands and procurement methods in the government and enterprise market. For example, policy adjustments on informatization construction by the government may affect the procurement scale and procurement direction of government informatization projects.

[0007] Combined with the above factors, the current TOB marketing strategies of most enterprises tend to focus on relationship marketing, and establishing and maintaining good relationships with government and enterprise customers is an important means of market promotion. However, for the continuous operation of the stock market, in order to achieve goals such as improving customer stickiness and product repurchase, relationship marketing will seem like overkill. More effective IT means are needed to complete the operation of the stock market and promote the continuous increase of the ARPU value in the stock market. Among them, using effective customer data for modeling and associating marketing strategies become the key means.

[0008] The solution "CN107451861B A method for identifying user Internet access characteristics under big data" describes that by analyzing data such as users' historical Internet access data, running trajectories, and residence durations, a complete unified view of guests is established, guest behavior attribute tags are constructed, and cross-marketing and targeted marketing for individuals are realized. However, its main marketing target is individual customers, and the collected data set is mainly completed by collecting information such as the login and logout times of users' Internet access, the links visited, and the page stay times. In the TOB marketing scenario, it does not have universality. Moreover, in this patent, the association relationship model between customers and products (or stores) only completes the product and customer portraits, and the relationship model cannot be applied to the business purchase and decision-making scenarios in the TOB scenario.

[0009] The solution "CN104182422B Unified address book information processing method and system" superimposes relationships such as contacts and friends between users into a unified network graph based on the address book, mines users' social relationships, and then promotes cross-marketing for businesses. However, although it pays attention to the transitivity of social relationships between customers, the customer group still remains in the TOC scenario, and the connection bond between customers is only address book information, which cannot effectively meet the problems of proxy decision-makers and long-process approval mechanisms in the TOB marketing and sales scenarios, nor can it solve the problem of associated marketing for TOB existing customers and improve the operation effect of government and enterprise existing customers.

[0010] In the solution "CN114118299A A clustering method combining similarity measurement and community discovery", the Sorensen-Dice index is used to calculate the similarity in the weight calculation between adjacent nodes. The calculation is based on the percentage calculation according to the index formula, and the calculation result of the algorithm is greatly affected by the number of attributes (scale), and the value range stays between [0, 1]. For the subsequent Louvain algorithm for community classification and node aggregation, the effect is not obvious. At the same time, in addition to the slightly poor clustering effect, the Sorensen-Dice index has insufficient stability in the community division effect. As the business evolves, when the number of attributes included in the calculation scope of the enterprise changes, but the number of equal attribute values (intersection) between nodes does not change, using the Sorensen-Dice index, the degree between nodes will decrease to varying degrees as the number of attributes increases, and the existing community division results between enterprises will change and adjust, resulting in a situation inconsistent with the actual business. Summary of the Invention

[0011] The purpose of the present invention is to provide a method and system for cross-marketing of government and enterprise customers based on graph analysis.

[0012] To solve the above technical problems, the present invention provides a method for cross-marketing of government and enterprise customers based on graph analysis, which is characterized by including the following steps:

[0013] Obtain basic data;

[0014] Construct a knowledge graph based on the basic data;

[0015] Perform community partitioning on the knowledge graph to obtain a number of communities;

[0016] Based on the average degree or similarity calculation, screen high-quality communities from all communities;

[0017] Recommend high-quality communities.

[0018] Preferably, constructing a knowledge graph based on the basic data specifically includes the following steps:

[0019] Extract entity, entity attribute information, and relationship information between entities from the basic data; the basic data includes enterprise information and product information;

[0020] Construct a knowledge graph according to the entity, entity attribute information, and relationship information between entities; the knowledge graph includes an enterprise knowledge graph and a product knowledge graph.

[0021] Preferably, performing community partitioning on the knowledge graph to obtain a number of communities specifically includes the following steps:

[0022] Obtain the edge weights between adjacent nodes in the knowledge graph;

[0023] Based on the Louvain algorithm, perform community partitioning on the knowledge graph according to the edge weights to obtain a number of communities; the communities include enterprise communities and product communities.

[0024] Preferably, calculating the edge weights of adjacent nodes in the knowledge graph specifically includes the following steps:

[0025] Obtain the edge weights between adjacent nodes according to the number of identical attributes between adjacent nodes in the knowledge graph and the corresponding weight coefficients.

[0026] Preferably, based on the average degree or similarity calculation, screening high-quality communities from all communities specifically includes the following three methods:

[0027] Method 1: Calculate the average degree of the community, and take the community with an average degree greater than the first threshold as a high-quality community;

[0028] Method 2: Calculate the average degree of the community, and according to the amount of common attributes and the average degree, take the community with a ratio of the amount of common attributes to the average degree within the community greater than the second threshold as a high-quality community;

[0029] Method 3: Calculate the target average similarity between nodes and target attributes in the community, and use the community where the similarity between the marketing target attribute and the node is greater than the third threshold as the high-quality community.

[0030] Preferably, the formula for calculating the average degree is:

[0031]

[0032] In the formula: The degree between any two nodes A and B in the community is the number of identical attributes between the two nodes.

[0033] Preferably, the formula for calculating the amount of common attributes is:

[0034]

[0035] In the formula: The number of common attributes in the community is the number of identical attributes of all nodes.

[0036] Preferably, the formula for calculating the target average similarity is:

[0037]

[0038] In the formula: A is the marketing strategy; i is the node in the community; n is the total number of nodes; J(i, A) is the similarity between node i and marketing strategy A.

[0039] Preferably, the recommendation for high-quality communities specifically includes the following steps:

[0040] Obtain nodes from high-quality communities as users to be recommended; the users to be recommended are enterprises or products to be recommended;

[0041] The users to be recommended obtain the minimum common attributes therefrom;

[0042] Compare the minimum common attributes with the preset exclusive attributes; if the minimum common attributes do not have the preset exclusive attributes, recommend the users to be recommended;

[0043] Compare the attributes of the users to be recommended with all the attributes of the corresponding high-quality community, and use the attributes that do not exist in the users to be recommended among all the attributes of the high-quality community as the recommended content.

[0044] The present invention also provides a system for cross-marketing of government and enterprise customers based on graph analysis, including:

[0045] An acquisition module for acquiring basic data;

[0046] A module for constructing a knowledge graph according to the basic data;

[0047] A community division module for dividing the knowledge graph into several communities;

[0048] A screening module for screening high-quality communities from all communities based on average degree or similarity calculation;

[0049] A recommendation module for recommending high-quality communities.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] Improve customer stickiness: Through cross-marketing, enterprises can provide more product and service combinations to government and enterprise customers to meet their diverse needs. For example, when providing an informatization system for government customers, relevant data security services, cloud computing resources, etc. can also be provided, making customers more dependent on the overall solution of the enterprise, thereby improving customer stickiness and loyalty.

[0052] Reduce marketing costs: Cross-marketing can make full use of existing customer resources and channel resources. Enterprises do not need to invest a large amount of marketing expenses to acquire new customers, but by providing more products and services to existing customers, they can achieve the efficient use of marketing resources and reduce marketing costs.

[0053] Enhance market competitiveness: Through cross-marketing, enterprises can expand the market coverage of products and services. For example, cross-promoting products and services in different fields such as IT, CT, and DT can better meet the comprehensive needs of government and enterprise customers in aspects such as informatization construction, communication network construction, and data processing, enhancing the market competitiveness of the enterprise.

[0054] Promote product innovation: Cross-marketing can promote enterprises to deeply understand and explore customer needs. In the process of providing multiple products and services to government and enterprise customers, enterprises can discover potential associations and synergy effects between different products, thereby promoting product innovation and developing more competitive new products and solutions.

[0055] Respond to market changes: In the current economic environment, the needs and competition pattern in the government and enterprise market are constantly changing. Through cross-marketing, enterprises can more flexibly adjust product portfolios and market strategies, quickly respond to market changes, seize new market opportunities, and achieve sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The following further elaborates the specific implementation manners of the present invention with reference to the drawings.

[0057] Figure 1 is a framework schematic diagram of a system for cross-marketing of government and enterprise customers based on graph analysis of the present invention;

[0058] Figure 2 is a schematic diagram of similarity calculation. DETAILED DESCRIPTION OF THE INVENTION

[0059] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific implementations disclosed below.

[0060] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a", "the", and "said" used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0061] It should be understood that although the terms first, second, etc. may be used in one or more embodiments of this specification to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein can be interpreted as "when" or "while" or "in response to determining".

[0062] The following further describes the present invention in detail with reference to the accompanying drawings:

[0063] The present invention provides a method for cross - marketing of government - enterprise customers based on graph analysis, which is characterized by including the following steps:

[0064] Obtain basic data;

[0065] Construct a knowledge graph based on the basic data;

[0066] Perform community partitioning on the knowledge graph to obtain several communities;

[0067] Screen high - quality communities from all communities based on average degree or similarity calculation;

[0068] Recommend the high - quality communities.

[0069] Preferably, constructing a knowledge graph based on the basic data specifically includes the following steps:

[0070] Extract entity, attribute information of the entity, and relationship information between entities from the basic data; the basic data includes enterprise information and product information;

[0071] Construct a knowledge graph based on entities, attribute information of entities, and relationship information between entities; the knowledge graph includes an enterprise knowledge graph and a product knowledge graph.

[0072] Preferably, perform community partitioning on the knowledge graph to obtain a number of communities, which specifically includes the following steps:

[0073] Obtain the edge weights between adjacent nodes in the knowledge graph;

[0074] Based on the Louvain algorithm, perform community partitioning on the knowledge graph according to the edge weights to obtain a number of communities; the communities include enterprise communities and product communities.

[0075] Preferably, calculate the edge weights of adjacent nodes in the knowledge graph, which specifically includes the following steps:

[0076] Obtain the edge weights between adjacent nodes according to the number of identical attributes between adjacent nodes in the knowledge graph and the corresponding weight coefficients.

[0077] Preferably, screen high-quality communities from all communities based on average degree or similarity calculation, which specifically includes the following three methods:

[0078] Method 1: Calculate the average degree of the community, and take the community with an average degree greater than the first threshold as a high-quality community;

[0079] Method 2: Calculate the average degree of the community, and take the community with a ratio of the amount of common attributes to the average degree within the community greater than the second threshold as a high-quality community according to the amount of common attributes and the average degree;

[0080] Method 3: Calculate the target average similarity between nodes in the community and the target attribute, and take the community with a similarity between the marketing target attribute and the nodes greater than the third threshold as a high-quality community.

[0081] Preferably, the calculation formula for the average degree is:

[0082]

[0083] In the formula: the degree between any two nodes AB in the community is the number of identical attributes between the two nodes.

[0084] Preferably, the calculation formula for the amount of common attributes is:

[0085]

[0086] In the formula: the number of common attributes in the community is the number of identical attributes of all nodes.

[0087] Preferably, the calculation formula for the target average similarity is:

[0088]

[0089] In the formula: A is the marketing strategy; i is the node in the community; n is the total number of nodes; J(i, A) is the similarity between node i and the marketing strategy A.

[0090] Preferably, for the recommendation of high-quality communities, the specific steps are as follows:

[0091] Obtain nodes from high-quality communities as users to be recommended; the users to be recommended are enterprises or products to be recommended;

[0092] The users to be recommended obtain the minimum common attributes therefrom;

[0093] Compare the minimum common attributes with the preset exclusive attributes; if the minimum common attributes do not have the preset exclusive attributes, recommend the users to be recommended;

[0094] Compare the attributes of the users to be recommended with all the attributes of the corresponding high-quality communities, and use the attributes that do not exist in the users to be recommended among all the attributes of the high-quality communities as the recommended content.

[0095] The present invention also provides a system for cross-marketing of government and enterprise customers based on graph analysis, including:

[0096] An acquisition module for acquiring basic data;

[0097] A module for constructing a knowledge graph according to the basic data;

[0098] A community division module for dividing the knowledge graph into several communities;

[0099] A screening module for screening high-quality communities from all communities based on the calculation of average degree or similarity;

[0100] A recommendation module for recommending high-quality communities.

[0101] The key points of the method for cross-marketing of government and enterprise customers based on graph analysis in the present invention are:

[0102] 1. Through the analysis method of graph and Louvain algorithm, complete the correlation analysis related to government and enterprise customers;

[0103] 2. Through the analysis method of graph and Louvain algorithm, complete the correlation analysis between products;

[0104] 3. Through the formed node communities of government and enterprise customer graphs and product graph node communities, cross-comparison is carried out to achieve cross-marketing.

[0105] In order to better illustrate the technical effects of the present invention, the following specific embodiments are provided to illustrate the above technical process:

[0106] Example 1. A method for cross - marketing of government and enterprise customers based on graph analysis. The main goal is to achieve cross - marketing of government and enterprise customers (TOB), which mainly consists of the following parts:

[0107] (1) Basic management capabilities, mainly for basic information management and maintenance

[0108] a) Enterprise basic information management: Through information such as the enterprise's internal CRM, synchronize the information of government and enterprise customers, and maintain the common attributes of the enterprise (industry, scale, type, business scope, intellectual property, etc.) for subsequent enterprise profiling and enterprise similarity analysis reference

[0109] b) Product management: Mainly complete the management of the available product list and product relationships. Among them, product list management mainly completes the list management of marketable / salable products. At the same time, for product relationships, focus on maintaining the degree of association between products under different market conditions. The degree of product association is calculated by the following "product association calculation engine" and automatically saved.

[0110] c) Customer order relationship management: Mainly through information such as enterprise internal project management and order management, complete the information management of the fulfilled orders of existing customers, and focus on recording basic information such as the customer and product order time, status, quantity, duration, etc.

[0111] d) Industry policy news / dynamic management: This module mainly supports methods such as manual input and public information crawling to complete the management of industry dynamics, providing input for whether to generate marketing activities in the future.

[0112] e) Marketing activity management: This module completes the definition of marketing activity plans. It mainly includes the definition and management of key information such as the activity object (enterprise characteristics, such as enterprise scale products, products already ordered by the enterprise, enterprise type, enterprise historical signing and ordering situation, etc.), activity start and end times, activity product scope, activity goals, etc. The management content serves as the input for enterprise similarity calculation tasks and product similarity calculation tasks to initiate relevant calculations and recommendations.

[0113] (2) Recommendation calculation capabilities, mainly for active marketing activity recommendations and similarity calculations based on customer and product similarities

[0114] a) Enterprise similarity calculation engine (community detection):

[0115] Detailed description of the enterprise similarity calculation process:

[0116] 1. Construct the graph nodes of the enterprise: The enterprise attribute information is managed and maintained through the module "Enterprise basic information management". And as the information changes, the relevant enterprise and attribute information is synchronously updated to the graph database (enterprise graph database).

[0117] 2. Calculate the edge weights between all adjacent nodes of the enterprise. The edge weight calculation method is to use the number of identical attributes between two enterprise nodes (AB) as the edge weight on the connection line between nodes AB. Considering the actual business needs, different weight coefficients are added to different attributes according to the differences of attributes (the default weight coefficient is the same for all attributes, which is 1). Taking the attribute weight of 1 as an example, as Figure 2 shown:

[0118] 3. Perform community division on the enterprise graph constructed in the above steps according to the Louvain algorithm. The specific steps refer to the algorithm steps. Finally, all enterprise customers will form different communities to achieve the classification of government and enterprises.

[0119] b) Product correlation calculation engine: The reference algorithm for product correlation is the Louvain algorithm. The specific process is as follows:

[0120] i. Construct the graph nodes of the product: The establishment process is the same as that of constructing the graph node type of enterprise customers. The maintenance of attributes is carried out through product management in the management function. Common product attributes such as product source, product type, product orderable customer range, product delivery method, and product marketing status can all be incorporated into the node construction process. And with the change of information, the relevant product and attribute information is synchronously updated to the graph database (product graph database).

[0121] ii. Calculate the degree between product graph nodes: The degree between product graph nodes is calculated by the number of attribute similarities between two products. Similarly, different weights are added to different attributes according to the differences of attributes (the default weight is the same for all attributes, which is 1). Specifically, according to the enterprise's product portfolio marketing strategy, the clustering weight of a certain type of attribute can be increased, such as attributes / labels like "enterprise-owned production equipment" and "products supplied by high-quality partners".

[0122] iii. Refer to the Louvain algorithm to perform community division on the product graph to form different product sets.

[0123] Marketing recommendations are made from two dimensions (different perspectives). One is starting from the enterprise, clustering the enterprise, forming an enterprise knowledge graph, finding similar customers, and "recommending" the products between customers to each other; the other recommendation dimension is starting from the product, finding related products (for example, customers who have ordered cloud hosts should also need cloud storage and cloud dedicated lines), and recommending related products to customers who have ordered a certain product.

[0124] The Louvain algorithm is a widely used community detection algorithm mainly for identifying community structures in complex networks. It was proposed by Vincent D. Blondel et al. in 2008 and is a greedy algorithm based on modularity optimization. The following is a detailed introduction to the Louvain algorithm:

[0125] Principle of the algorithm

[0126] Modularity: Modularity is an important indicator to measure the quality of community division. It reflects the difference between the connection density of nodes within a community in the network and the expected connection density in a random network. The calculation formula for modularity is:

[0127]

[0128] where: Aij is the edge weight between node i and node j, ki and kj are the degrees of node i and node j respectively, m is the sum of the weights of all edges in the graph, and δ(ci, cj) is an indicator function that is 1 when node i and node j belong to the same community and 0 otherwise.

[0129] Steps of the algorithm:

[0130] 1. Initialization: Consider each node as an independent community, and the community label of each node is its own node number.

[0131] 2. Local optimization: For each node, traverse all its neighbor nodes and calculate the increase in modularity when adding this node to the community where the neighbor node is located. Select the neighbor node with the largest increase in modularity and add this node to its community. This process is repeated until no node can increase the modularity by changing the community label.

[0132] 3. Network compression: Consider each community as a node, convert the weights of nodes within the community into self-loop weights of the new node, and convert the weights of nodes between communities into edge weights between the new nodes.

[0133] 4. Iteration: Repeat the above steps of local optimization and network compression until the modularity no longer increases.

[0134] (3) Operation and effect evaluation, mainly completing cross-marketing recommendations for government and enterprise customers, marketing effect evaluation, and model iteration to form a complete closed loop

[0135] a) Cross-marketing recommendations for government and enterprise customers are mainly divided into two parts: mutual recommendations of different subscribed products for the same government and enterprise customer community, and mutual recommendations of different customers for the same product community:

[0136] i. Different product recommendations for the same customer community: Find the list of all ordered products P1 in the same government and enterprise customer community C1, compare all customer nodes in this community, and for all unordered products, combine the screening of marketing strategies (maintain strategies through the "marketing activity management" function, that is, which products can be sold to what types of customers through what touchpoints and at what times), complete the generation of cross-recommendation tasks for customer nodes in this community, and distribute them to the corresponding touchpoints according to the marketing strategy. Specifically, considering the factor of the scale of the enterprise customer group, complete the screening for the marketing strategy through multi-level screening:

[0137] Screen high-quality customer communities: Modularity can only be used to judge the link density within a community. After community division based on the Louvain algorithm in combination with the cross-marketing success rate, modularity alone cannot effectively measure the relevance of nodes within a community from a business perspective. The division of customer communities depends on the degree between nodes AB, which is the equal number of attributes. There will be a certain deviation in the degrees between multiple nodes in the same community, and there will also be a large deviation in the average degree of multiple nodes between different communities. Considering the marketing effect, the higher the similarity of community business, the greater the cross-marketing success rate. The main ways to specifically judge the similarity of community business are as follows:

[0138] Method 1: Judge the average degree situation within the community based on a preset threshold. Screen high-quality customer communities through the average degree threshold of the community, that is, if the average degree of community nodes > the first threshold, it is included in the category of high-quality communities, otherwise it is eliminated. The average degree f(x) of community nodes refers to:

[0139]

[0140] Method 2: Use the ratio of the amount of common attributes to the average degree within the community as a basis for screening high-quality communities for community business. The higher the ratio of the amount of common attributes to the average degree within the community, the higher the probability that nodes within the current community are assigned to the same community through the same attributes; the amount of common attributes within the community is the degree between all nodes within the community;

[0141]

[0142] If the ratio of the amount of common attributes to the average degree within the community > the second threshold, it is included in the category of high-quality communities, otherwise it is eliminated.

[0143] Method 3: Consider that the main purpose of screening high-quality customer communities is to recommend targeted products to customers close to the marketing target. In order to screen out customer communities close to the marketing target, conduct a similarity analysis of the marketing target attributes and nodes in the community (average similarity between the community and the marketing target):

[0144]

[0145] Formula represents the calculation of the similarity between node i and marketing strategy A through the Jaccard coefficient

[0146] If the similarity between the marketing target attribute and the node is greater than the third threshold, it is included in the category of high-quality communities; otherwise, it is eliminated.

[0147] Screen the set of minimum common attributes: Find all the common attributes of all nodes from the enterprise communities to be recommended.

[0148] Marketing strategy matching: Marketing strategies are divided into recommendation and exclusive attribute settings. The key to strategy matching is the screening between exclusive attributes and the set of minimum common attributes. If there are exclusive attributes in the current minimum attribute set, the current community skips the recommendation and proceeds to the next community for judgment;

[0149] Product matching: After the above screening, it enters the product comparison link. Compare the products already ordered by node C1 in the community with the set of recommended products P1. If not, it enters the recommendation list.

[0150] ii. Recommendation for different customers in the same product community: Similar to the recommendation of different products in the same customer community, first screen out all valid government and enterprise customer lists in the same product community, compare all products already ordered by the current customers with all products in the current product community, and at the same time combine the screening of marketing strategies. For the situation that meets the marketing strategy requirements and is not ordered by the customer in the current product community, complete the generation of cross-recommendation tasks and send them to the corresponding touchpoints in combination with the marketing strategy. For the specific screening strategy, refer to the customer's screening process for products, eliminate product communities that do not meet the marketing requirements from screening high-quality product communities, screening the set of minimum common attributes, and comparing marketing strategies, and finally perform node matching to complete cross-marketing for different customers of the same product.

[0151] b) Marketing effect evaluation: Combine the information reception, product browsing, and ordering of cross-marketing recommendation tasks on different touchpoints by customers to complete the collection, summary, and analysis of marketing results, form an analysis of marketing effects (success rates) and a comparison of historical effects, and form an iterative feedback from the marketing activities, enterprise similarity, and product correlation calculation models.

[0152] c) Model iteration: Combine the feedback and focus on optimizing the relevant models for calculating enterprise similarity and product correlation in the recommendation calculation ability.

[0153] Attribute Optimization: Focus on completing the screening and elimination of attributes relied on by enterprises and the screening and elimination of attributes related to products, and select efficient attributes as the basis for degree calculation. Example: In the calculation of product communities, the product's listing time may gradually become a low-quality attribute for customer orders over time. In the calculation of government and enterprise customer communities, as the life cycle of government and enterprise customers changes, the role of the bank where government and enterprise customers open their accounts becomes weaker and will become an attribute to be eliminated.

[0154] i. Attribute Weight Optimization: Combine the actual business requirements to adjust the attribute weights of different attributes of customers / products;

[0155] ii. High-quality Community Threshold Optimization: Combine the scope of marketing, adjust the business strategy, and optimize the threshold in combination with the marketing success rate.

[0156] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules, modules or units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units, modules or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0157] The unit may or may not be physically separated. The components shown as units may be a physical unit or multiple physical units, that is, they may be located in one place, or may be distributed to multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0158] In addition, the functional units in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0159] In particular, according to the embodiments disclosed in the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), the above-mentioned functions defined in the methods of the present invention are performed. It should be noted that the above-mentioned computer-readable medium of the present invention can be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above.

[0160] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0161] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for cross-marketing of government and enterprise customers based on graph analysis, characterized in that: The following steps are involved: Get basic data; Build a knowledge graph based on basic data; Divide the knowledge graph into communities and obtain several communities; Filter high-quality communities from all communities based on average or similarity calculation; Recommend high-quality communities.

2. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 1 is characterized in that: Based on the basic data, the knowledge graph is constructed, which includes the following steps: Extracting entities, attribute information of entities, and relationship information between entities from basic data; the basic data includes enterprise information and product information; A knowledge graph is constructed based on entities, attribute information of entities, and relationship information between entities; the knowledge graph includes an enterprise knowledge graph and a product knowledge graph.

3. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 2 is characterized in that: Divide the knowledge graph into communities and obtain several communities, which specifically includes the following steps: Get the edge weights between adjacent nodes in the knowledge graph; Based on the Louvain algorithm, the knowledge graph is divided into communities according to edge weights to obtain several communities; the communities include enterprise communities and product communities.

4. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 3 is characterized in that: Calculating the edge weights of adjacent nodes in the knowledge graph includes the following steps: According to the number of the same attributes between adjacent nodes in the knowledge graph and the corresponding weight coefficients, the edge weights between the adjacent nodes are obtained.

5. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 4 is characterized in that: Based on average or similarity calculation, select high-quality communities from all communities, including the following three methods: Method 1: Calculate the average degree of the community, and take the community with an average degree greater than the first threshold as a high-quality community; Method 2: Calculate the average degree of the community, and based on the common attribute quantity and the average degree, take the community whose common attribute quantity and average degree ratio is greater than the second threshold as a high-quality community; Method three: Calculate the target average similarity between nodes and target attributes in the community, and take the community whose similarity between marketing target attributes and nodes is greater than the third threshold as a high-quality community.

6. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 5 is characterized in that: The calculation formula of the average degree is: Where: The degree of a community between any two nodes AB is the number of identical attributes between the two nodes.

7. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 6 is characterized in that: The calculation formula of the common attribute quantity is: Where: The number of common attributes in the community is the number of the same attributes of all nodes.

8. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 7 is characterized in that: The calculation formula of the target average similarity is: Where: A is the marketing strategy; i is the node in the community; n is the total number of nodes; J(i, A) is the similarity between node i and marketing strategy A.

9. The method for cross-marketing of government and enterprise customers based on graph analysis according to claim 8 is characterized in that: Recommending high-quality communities includes the following steps: Obtain nodes from high-quality communities as users to be recommended; users to be recommended are companies or products to be recommended; The minimum common attributes are obtained from the users to be recommended; Compare the minimum common attribute with the preset exclusive attribute; if the minimum common attribute does not have the preset exclusive attribute, recommend the user to be recommended; The attributes of the user to be recommended are compared with all the attributes of the corresponding high-quality community, and the attributes of all the attributes of the high-quality community that do not exist in the user to be recommended are used as recommendation content.

10. A system for cross-marketing of government and enterprise customers based on graph analysis, used to implement the method for cross-marketing of government and enterprise customers based on graph analysis as described in any one of claims 1 to 9, characterized in that: include: Acquisition module, used to obtain basic data; Module, used to build knowledge graph based on basic data; The community division module is used to divide the knowledge graph into communities and obtain several communities; The screening module is used to screen high-quality communities from all communities based on average degree or similarity calculation; The recommendation module is used to recommend high-quality communities.

Images