Screening method of potential corporate customers of banks based on industrial chain map
Generate directed graphs through the industrial chain map method, and use similarity and connection paths to screen banks' corporate customers, solving the problem of low screening efficiency and accuracy in the existing technology, and achieving efficient and accurate potential customer identification.
Patent Information
- Application Number
- CN202111401213.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-19
- Publication Date
- 2024-12-24
- Estimated Expiration
- 2041-11-19
AI Technical Summary
In the prior art, in bank-based corporate customer screening, the rule-driven method has low accuracy and poor timeliness. The machine learning method requires a large number of samples, resulting in low screening efficiency and accuracy.
Based on the industrial chain map, by generating directed graphs, using the similarity and connection paths of core customers, we quickly filter out potential customers related to core customers without a large number of samples and rules.
It improves the efficiency and accuracy of potential customers, can accurately identify potential customers with strong correlation with core customers, and reduces the complexity of the screening process.
Smart Images
Figure CN114372815B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a method for screening potential corporate customers of a bank based on an industrial chain map. Background Art
[0002] Corporate business, as the most important asset and source of income for banks, has a slower digital upgrade process than individual retail business. Currently, corporate products of banks are highly homogenized, and the competition in corporate business is becoming increasingly fierce. As the country promotes inclusive finance and policies focus on supporting small and medium-sized enterprises, banks will face a huge number of corporate customer groups. Therefore, how to effectively screen out potential customers from the massive corporate customer groups is the key to improving the efficiency of corporate marketing.
[0003] In related technologies, rule-driven methods can be used to describe specific scenarios and target customer groups in the form of rules to form score cards or customer portraits, thereby screening out potential customers based on the score cards or customer portraits. Alternatively, machine learning can be used to mine the bank's massive data, thereby conducting a comprehensive analysis of business scenarios and target customer groups to screen out potential customers.
[0004] However, since the rule-driven approach is limited by business experience, not all fields have good experience summaries. Moreover, weak rules or rules composed of fuzzy conditions often have low accuracy. At the same time, rules may conflict with each other, resulting in inaccurate screening of potential customers. Secondly, the rules are time-sensitive and need to be adjusted regularly, resulting in low efficiency in screening corporate customers. Using machine learning to mine the massive data of banks requires a sufficient number of samples to reflect the full picture of things, but most commercial banks do not have sufficient sample annotations, which increases the difficulty of applying supervised learning methods, which in turn reduces the accuracy of screening potential customers. Summary of the invention
[0005] The embodiment of the present application provides a method for screening potential corporate customers of a bank based on an industrial chain map, which improves the efficiency of identifying potential customers while improving the accuracy of the screened potential customers.
[0006] In a first aspect, the present application embodiment provides a method for screening potential corporate customers of a bank based on an industrial chain map, comprising:
[0007] Generate nodes corresponding to each customer based on their account information;
[0008] A connection path is formed between two nodes corresponding to any two customers in the same industrial chain who are in an upstream and downstream relationship and conduct transactions within a preset period of time, so as to generate a directed graph;
[0009] Performing similarity matching between each first node connected to the core node corresponding to the core customer in the directed graph and the core node to determine the similarity between each first node and the core node, and performing similarity matching between each second node connected to the same node as the core node in the directed graph and the core node to determine the similarity between each second node and the core node;
[0010] According to each of the similarities, at least one target node is determined from each of the first nodes and each of the second nodes, so as to determine the customer corresponding to the target node as a potential customer.
[0011] In one embodiment, each first node connected to a core node corresponding to a core customer in the directed graph is matched with the core node for similarity, and the similarity between each first node and the core node is determined, including:
[0012] Determining the similarity between each of the first nodes and the core nodes corresponding to the core customers according to the transaction scale ratios of each of the target customers and the core customers corresponding to each of the first nodes;
[0013] The transaction scale ratio is the ratio of the net transaction value between the core customers and the target customers to the sum of all net transaction values generated by the core customers.
[0014] In one embodiment, the net transaction value is determined based on the transaction difference between the target customer and the core customer within the preset time period and a preset time weight.
[0015] In one embodiment, performing similarity matching between each second node in the directed graph that is connected to the same node as the core node and the core node, and determining the similarity between each second node and the core node, includes:
[0016] Determining the similarity between each of the second nodes and the core node according to the proportion of the same connection paths between each of the second nodes and the core node;
[0017] The proportion of the same connection paths is the ratio of the total number of common inflow and outflow connection paths between the second node and the core node in the directed graph to the total number of connection paths between the second node and the core node in the directed graph respectively inflow and outflow.
[0018] In one embodiment, the upstream and downstream relationships between the customers are determined based on the upstream and downstream relationships between the industries corresponding to the customers.
[0019] In one embodiment, determining at least one target node from each of the first nodes and each of the second nodes according to each of the similarities includes:
[0020] Filter out, from each of the second nodes, each candidate node whose similarity to the second node is greater than a preset threshold;
[0021] At least one target node is determined from each of the first nodes and each of the candidate nodes according to the similarity between each of the first nodes and each of the candidate nodes and the core node.
[0022] In one embodiment, determining at least one target node from each of the first nodes and each of the candidate nodes according to the similarity between each of the first nodes and each of the candidate nodes and the core node includes:
[0023] Connecting each of the candidate nodes to the core node;
[0024] According to the similarity between each of the first nodes and each of the candidate nodes and the core node, the attributes of the connection paths between each of the first nodes and each of the candidate nodes and the core node are updated to obtain a directed weighted graph;
[0025] Performing operations on the directed weighted graph according to the PageRank webpage ranking algorithm to obtain the node weights of each of the first nodes and each of the second nodes to be selected in the directed weighted graph;
[0026] According to the weights of the nodes, at least one target node is determined from the first nodes and the nodes to be selected, so as to determine the customer corresponding to the target node as a potential customer.
[0027] In one embodiment, determining at least one target node from each of the first nodes and each of the candidate nodes according to the weight of each of the nodes, so as to determine the customer corresponding to the target node as a potential customer, includes:
[0028] Sort the first nodes and the nodes to be selected according to their node weights, and obtain rankings of the first nodes and the nodes to be selected;
[0029] From each of the first nodes and each of the nodes to be selected, each of the target nodes ranked before a preset ranking is obtained, so as to determine each of the customers corresponding to each of the target nodes as a potential customer.
[0030] In one embodiment, the attribute of the connection path is transaction summary information between two customers corresponding to the two connected nodes.
[0031] In one embodiment, the direction of the connection path is determined according to the upstream and downstream relationship between two clients corresponding to the two nodes.
[0032] The method for screening potential corporate customers of a bank based on an industrial chain map provided in an embodiment of the present application generates a directed graph based on customers who are upstream and downstream of each other in the same industrial chain, and takes the core node corresponding to the core customer as the starting point, and uses the connection path in the directed graph to obtain each first node that is directly upstream and downstream of the core node, and each second node that is connected to the same node as the core node, that is, has a common upstream and downstream node with the core node, and quantifies the relationship between each second node and each second node and the core node according to the similarity between each second node and the core node, so that the connection path and similarity can be used to quickly find customers with strong correlation with the core customer corresponding to the core node as potential customers, and potential customers can be screened out without providing a large number of samples and rules, thereby improving the efficiency of potential customer determination. And using the connection path and similarity between the nodes of the directed graph, potential customers with strong correlation with the core customer can be accurately determined, thereby improving the accuracy of the potential customers screened out. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0034] Figure 1 It is a flow chart of a method for screening potential corporate customers of a bank based on an industrial chain map provided in an embodiment of the present application;
[0035] Figure 2 is a schematic diagram of a connection structure between a first node and a core node provided in an embodiment of the present application;
[0036] Figure 3 It is a schematic diagram of the connection structure between the second node and the core node provided in an embodiment of the present application. DETAILED DESCRIPTION
[0037] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0038] The embodiments of the present application are described in detail below with reference to the accompanying drawings.
[0039] See also Figure 1, is one of the flow diagrams of a method for screening potential corporate customers of a bank based on an industrial chain map provided by an embodiment of the present invention. The method is applied to an electronic device, wherein the electronic device may be a server or a terminal device, and is used to identify potential customers from a large number of corporate customers. Figure 1 As shown, the present embodiment provides a method for screening potential corporate customers of a bank based on an industrial chain map, including:
[0040] Step 101, generating nodes corresponding to each customer one by one according to the account information of each customer;
[0041] Step 102, forming a connection path between two nodes corresponding to any two customers in the same industrial chain and in an upstream and downstream relationship and conducting transactions within a preset period of time, so as to generate a directed graph;
[0042] Step 103, performing similarity matching between each first node connected to the core node corresponding to the core customer in the directed graph and the core node, and determining the similarity between each first node and the core node, and performing similarity matching between each second node connected to the same node as the core node in the directed graph and the core node, and determining the similarity between each second node and the core node;
[0043] Step 104: Determine at least one target node from each of the first nodes and each of the second nodes according to each of the similarities, so as to determine the customer corresponding to the target node as a potential customer.
[0044] By generating a directed graph based on customers who are upstream and downstream of each other in the same industrial chain, and taking the core node corresponding to the core customer as the starting point, using the connection path in the directed graph, obtaining each first node that is directly upstream and downstream of the core node, and each second node that is connected to the same node as the core node, that is, having a common upstream and downstream node with the core node, and quantifying the relationship between each second node and each second node and the core node according to the similarity between each second node and the core node, it is possible to use the connection path and similarity to quickly find customers with strong correlation with the core customer corresponding to the core node as potential customers, and potential customers can be screened out without providing a large number of samples and rules, thereby improving the efficiency of potential customer determination. And using the connection path and similarity between the nodes of the directed graph, it is possible to accurately determine potential customers with strong correlation with the core customer, thereby improving the accuracy of the potential customers screened out.
[0045] In step 101, the customer's account information may be bank public account information, which includes customer identification, the industry to which the customer belongs, related business, the customer's historical transaction records, and transaction scale, etc. Then, the customer's account information is generated into a corresponding node, so that the node is used to represent the customer's account information.
[0046] In one embodiment, each customer is a customer in the same industrial chain.
[0047] In step 102, after generating each node representing the account information of each customer, a connected graph G=(V,E) representing the upstream and downstream relationship between each customer is constructed according to each node generated corresponding to each customer in the same industrial chain. Wherein, V represents the set of each node, E represents the edge set, and an edge represents the connection path between two nodes. The attribute of the connection path between any two connected nodes is the transaction summary information between the two customers corresponding to the two connected nodes. Wherein, the transaction summary information can specifically be the net transaction value.
[0048] In one embodiment, the direction of the connection path is determined according to the upstream and downstream relationship between two clients corresponding to the two nodes.
[0049] In one embodiment, the upstream and downstream relationships between the customers are determined based on the upstream and downstream relationships between the industries corresponding to the customers.
[0050] In one embodiment, whether any two industries in the same industrial chain are in an upstream-downstream relationship can be determined by judging the industry correlation and transaction direction between the two industries.
[0051] In one embodiment, the level of industry correlation between any two industries can be determined based on the transaction density between the two industries. Specifically, when the ratio of the total number of customer pairs with transactions between the two industries among all customers is greater than a preset first threshold, the correlation between the two industries is determined to be high. Two customers with transactions constitute a customer pair. The first threshold can be set according to actual conditions.
[0052] In one embodiment, whether the transaction directions of any two industries are concentrated can be determined based on the transaction direction concentration of the two industries. Specifically, if there are transactions between customers in two industries, then one of the directions with more transactions is selected as the industry transaction direction based on whether the net transaction amount is greater than 0 or less than 0, and the ratio of the number of customer pairs in this direction to the total number of customer pairs with transactions is determined as the transaction direction concentration. When the transaction direction concentration is greater than a preset second threshold, it is determined that the transaction directions between the two industries are concentrated. The second threshold can be set according to actual conditions.
[0053] When the correlation between two industries is high and the transaction direction is concentrated, it is determined that there is an upstream and downstream relationship between the two industries, and the upstream and downstream directions are the transaction directions.
[0054] In one embodiment, each customer is preset with an industry label to which it belongs. After determining the upstream and downstream relationships between industries, the industry to which each customer belongs can be identified based on the industry label to which each customer belongs, thereby determining the upstream and downstream relationships between customers based on the upstream and downstream relationships between industries.
[0055] In one embodiment, a time window T is preset, such as a preset period of three months, six months or one year. If two customers have transactions within the time window T and are upstream and downstream in the industrial chain, a directed edge from upstream to downstream is established between the two nodes corresponding to the two customers. The edge attribute is the net transaction value, which may be negative depending on the direction.
[0056] For example, customer A is upstream of customer B in the industrial chain. At the same time, customer A and customer B have a transaction within the time serial port T. Then, a directed edge from node 1 to node 2 is established between node 1 corresponding to customer A and node 2 corresponding to customer B.
[0057] After connecting all the nodes according to the above situation, we can get a directed graph G = (V, E).
[0058] In step 103, after obtaining the directed graph, based on the given core customer list, the core nodes U corresponding to the core customers are c As the source point, find the core node U in the directed graph c connected, that is, connected to the core node U c The first nodes U that are mutually upstream and downstream p .U c U p Represents a given core node U c The corresponding core customer and the first node U p The corresponding target customers have direct transactions within the time window T, such as Figure 2 After finding each first node, the similarity between each first node and the core node is matched to determine the similarity between each first node and the core node.
[0059] In order to make the obtained similarity more accurate, in one embodiment, the relationship between the first node and the core node can be quantified to determine the similarity between the first node and the core node. Specifically, the similarity between each first node and the core node corresponding to the core customer is determined based on the transaction scale ratio between each target customer and the core customer corresponding to each first node;
[0060] The transaction scale ratio is the ratio of the net transaction value between the core customers and the target customers to the sum of all net transaction values generated by the core customers.
[0061] In one embodiment, the net transaction value is determined based on the transaction difference between the target customer and the core customer within the preset time period and a preset time weight.
[0062] For example, assuming that the core customer is i, the target customer is j, and the time window is T, the net transaction value between the core customer and the target customer can be estimated based on the transaction flow: for:
[0063]
[0064] The transaction difference is the value of the transaction. A transaction difference greater than 0 indicates capital outflow, and a transaction difference less than 0 indicates capital inflow. ω(t) is the preset weight for time, and a time decay function is usually used.
[0065] After determining the net transaction value between core customers and target customers, the strength of the relationship between the two can be determined based on the proportion of the transaction scale between target customers and core customers, and the strength of the relationship can be used as the similarity between the first node corresponding to the target customer and the core node corresponding to the core customer.
[0066] Among them, the transaction scale of target customers and core customers accounts for:
[0067]
[0068] Among them, the numerator represents the net transaction value between core customer i and target customer j, and the denominator represents the sum of the net transaction values of all core customer i transactions; k represents the customer who has transactions with the core customer. Since the smaller the transaction scale of the target customer to the core customer, the weaker the industry relationship between the upstream or downstream target customer and the core customer, and vice versa, the stronger the industry relationship. Therefore, the proportion of transaction scale can be As the first node corresponding to target customer j, the similarity with the core node corresponding to core customer i Thereby, the similarity between the first node and the core node is determined accurately and effectively.
[0069] At the same time, the core node U corresponding to the core customer c As the source point, find the core node U in the directed graph c Connect to the same node U m , that is, with the core node U c Have a common upstream node or downstream node U m Each second node U p ′. c U m U p ′ represents a given core node U c The corresponding core customer and the second node U pThe corresponding customers have the same upstream or downstream suppliers. Figure 3 After finding each second node, each second node is matched with the core node to determine the similarity between each second node and the core node.
[0070] Since there is no direct transaction between the customers corresponding to the second nodes and the core customers corresponding to the core nodes, the relationship between the second nodes and the core nodes can be quantified. The similarity between each second node and the core node can be determined based on the proportion of the same connection paths between each second node and the core node; wherein the proportion of the same connection paths is the ratio of the total number of common inflow and outflow connection paths between the second node and the core node in the directed graph to the total number of inflow and outflow connection paths between the second node and the core node in the directed graph, respectively.
[0071] Exemplarily, after obtaining each second node, the strength of the relationship between the second node and the core node can be determined based on the common associated nodes of the second node and the core node, that is, the proportion of the same connection path between the second node and the core node is:
[0072]
[0073] The numerator represents the total number of common inflow or outflow paths between the core node i and the second node j, which is determined by the total number of common upstream nodes and common downstream nodes between the core node i and the second node j. The denominator represents the total number of all inflow and outflow paths between the core node i and the second node j. e′ ij Indicates whether there is a binary relationship with a directed edge from core node i to second node j, which is defined as follows:
[0074]
[0075] The smaller the ratio of the total number of common inflow or outflow paths between the core node i and the second node j to the total number of all inflow and outflow paths between the core node i and the second node j, the less intersection there is between the two industries corresponding to the two nodes, that is, the weaker the industry relationship between the two industries corresponding to the two nodes, and vice versa. Therefore, the proportion of the same connection path can be As the second node j, the similarity with the core node i Thereby, the similarity between the second node and the core node is accurately and effectively determined.
[0076] In one embodiment, after determining the similarity between each first node and each second node and the core node, the first node or second node corresponding to the similarity higher than the preset value may be used as the target node, and the customer corresponding to the target node may be determined as a potential customer. The preset value may be set according to actual conditions.
[0077] Considering that it is usually difficult to set the preset value, the first nodes and the second nodes can also be sorted according to the similarity from large to small, and the nodes ranked in the top N positions are obtained as target nodes, where N is a positive integer.
[0078] However, since each second node is not a node directly connected to the core node, that is, it is not directly related to the core node, while each first node is directly related to the core node. Therefore, in reality, even if the similarity between the second node and the core node is greater than the similarity between the first node and the core node, the correlation between the first node and the core node may still be greater than the correlation between the second node and the core node. At this time, if the target node is directly sorted and determined based on the similarity between each first node and each second node and the core node, it may happen that the users corresponding to some target nodes are not potential users in the actual sense, resulting in inaccurate screening results. To this end, in one embodiment, determining at least one target node from each first node and each second node based on each similarity includes:
[0079] Filter out, from each of the second nodes, each candidate node whose similarity to the second node is greater than a preset threshold;
[0080] At least one target node is determined from each of the first nodes and each of the candidate nodes according to the similarity between each of the first nodes and each of the candidate nodes and the core node.
[0081] In one embodiment, after obtaining the similarity between each second node and the core node, according to a pre-set preset threshold, each candidate node with a similarity greater than the preset threshold is first screened out from the second nodes, and then each candidate node and each first node are screened according to the similarity ranking, thereby screening out N target nodes.
[0082] By preliminarily screening each second node according to a preset threshold, the correlation between each selected node and the core node is sufficiently high, so that the correlation between the selected node and the core node is not weaker than the correlation between the first node and the core node. At this time, the target node is screened out from each candidate node and each first node according to the similarity sorting, so that the correlation between the selected target node and the core node is also sufficiently high, thereby making the potential users determined according to the target node more in line with actual needs and improving the accuracy of the screened potential users.
[0083] To further improve the accuracy of the potential users screened out, in one embodiment, according to the similarity between each of the first nodes and each of the candidate nodes and the core node, at least one target node is determined from each of the first nodes and each of the candidate nodes, including:
[0084] Connecting each of the candidate nodes to the core node;
[0085] According to the similarity between each of the first nodes and each of the candidate nodes and the core node, the attributes of the connection paths between each of the first nodes and each of the candidate nodes and the core node are updated to obtain a directed weighted graph;
[0086] Performing operations on the directed weighted graph according to the PageRank webpage ranking algorithm to obtain the node weights of each of the first nodes and each of the second nodes to be selected in the directed weighted graph;
[0087] According to the weights of the nodes, at least one target node is determined from the first nodes and the nodes to be selected, so as to determine the customer corresponding to the target node as a potential customer.
[0088] In one embodiment, after each candidate node is screened, a directed connection path from the core node to the candidate node is generated between each candidate node and the core node to update the directed graph, and a potential customer subgraph consisting of the core node and each first node and each candidate node with a connection path thereto is extracted from the updated directed graph, and then the attribute of each connection path in the potential customer subgraph is updated to the similarity between the core node at both ends of the path and the first node or the candidate node, thereby obtaining a directed weighted graph G. ′ =(V ′ ,E ′ ). Where V represents a node subset consisting of core nodes, first nodes, and candidate nodes, E ′ Represents a directed edge, that is, a collection of connection paths, and the edge attribute represents the similarity obtained by quantifying the relationship.
[0089] After obtaining the directed weighted graph, the directed weighted graph is operated through the PageRank web page ranking algorithm, specifically:
[0090] Each first node and each candidate node are regarded as first-degree neighbor nodes, and the initial weight of the core node in the directed weighted graph is a0. (1) Set to 1, the initial weight of each first-degree neighbor node is a0 (2) Set to 0 for initialization.
[0091] After initialization, iterative recursive optimization is performed. The core point is used as the current starting point. Starting from the current starting point, its weight is transferred to the corresponding first-degree neighbor nodes according to the similarity corresponding to each connection path, and the node weights of the core node and each first-degree neighbor node are updated in different ways. Then, the first-degree neighbor node and the current starting point set are merged as the starting point of the next round of iteration, and the node weights of each first-degree neighbor node around the starting point are continued to be updated.
[0092] For example, the i-th step can accept objects in two ways according to different node weights. Among them, the node weight of the core node v can be updated as:
[0093]
[0094] The node weight of the first-degree neighbor node v′ can be updated as:
[0095]
[0096] Where D is the damping coefficient, Represents the similarity between two nodes.
[0097] When the difference between the updated node weight and the node weight before the update is less than a certain set threshold, the iteration is stopped. The set threshold can be set according to the actual situation.
[0098] After the iteration stops, the node weights of each first-degree neighbor node are output, and at least one first-degree neighbor node can be selected as the target node according to the node weights of each first-degree neighbor node. For example, a first-degree neighbor node whose node weight is greater than a preset weight is selected as the target node, so that the customer corresponding to the target node is determined as a potential customer.
[0099] Considering that the setting of the preset weights needs to rely on experience, the setting is difficult. Therefore, in one embodiment, the first nodes and the candidate nodes may be sorted according to their node weights to obtain the rankings of the first nodes and the candidate nodes;
[0100] From each of the first nodes and each of the nodes to be selected, each of the target nodes ranked before a preset ranking is obtained, so as to determine each of the customers corresponding to each of the target nodes as a potential customer.
[0101] In one embodiment, the first-degree neighbor nodes composed of the first nodes and the candidate nodes are sorted from high to low according to the node weights of the first-degree neighbor nodes, and then the first-degree neighbor nodes ranked in the top N positions are extracted as target nodes, so as to determine the user corresponding to the target node as a potential user, thereby eliminating the need to set preset weights. This avoids the situation where too many or too few target nodes are extracted, or even no target nodes are extracted, resulting in inaccurate screening of potential users due to inaccurate preset weight settings, thereby improving the screening efficiency for potential users.
[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for screening potential corporate customers of a bank based on an industrial chain map, characterized in that: include: Generate nodes corresponding to each customer based on their account information; A connection path is formed between two nodes corresponding to any two customers in the same industrial chain who are in an upstream and downstream relationship and conduct transactions within a preset period of time, so as to generate a directed graph; Performing similarity matching between each first node connected to the core node corresponding to the core customer in the directed graph and the core node to determine the similarity between each first node and the core node, and performing similarity matching between each second node connected to the same node as the core node in the directed graph and the core node to determine the similarity between each second node and the core node; According to each of the similarities, at least one target node is determined from each of the first nodes and each of the second nodes, so as to determine the customer corresponding to the target node as a potential customer.
2. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 1, characterized in that: Performing similarity matching between each first node connected to the core node corresponding to the core customer in the directed graph and the core node, and determining the similarity between each first node and the core node, includes: Determining the similarity between each of the first nodes and the core nodes corresponding to the core customers according to the transaction scale ratios of each of the target customers and the core customers corresponding to each of the first nodes; The transaction scale ratio is the ratio of the net transaction value between the core customers and the target customers to the sum of all net transaction values generated by the core customers.
3. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 2 is characterized in that: The net transaction value is determined according to the transaction difference between the target customer and the core customer within the preset time period, and a preset time weight.
4. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 1 or 2, characterized in that: The method further comprises: performing similarity matching between each second node in the directed graph that is connected to the same node as the core node and the core node, and determining the similarity between each second node and the core node, including: Determining the similarity between each of the second nodes and the core node according to the proportion of the same connection paths between each of the second nodes and the core node; The proportion of the same connection paths is the ratio of the total number of common inflow and outflow connection paths between the second node and the core node in the directed graph to the total number of connection paths between the second node and the core node in the directed graph respectively inflow and outflow.
5. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 1, characterized in that: The upstream and downstream relationships between the customers are determined based on the upstream and downstream relationships between the industries corresponding to the customers.
6. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 1, characterized in that: The determining at least one target node from each of the first nodes and each of the second nodes according to each of the similarities comprises: Filter out, from each of the second nodes, each candidate node whose similarity to the second node is greater than a preset threshold; At least one target node is determined from each of the first nodes and each of the candidate nodes according to the similarity between each of the first nodes and each of the candidate nodes and the core node.
7. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 6, characterized in that: Determining at least one target node from each of the first nodes and each of the candidate nodes according to the similarity between each of the first nodes and each of the candidate nodes and the core node, including: Connecting each of the candidate nodes to the core node; According to the similarity between each of the first nodes and each of the candidate nodes and the core node, the attributes of the connection paths between each of the first nodes and each of the candidate nodes and the core node are updated to obtain a directed weighted graph; Performing operations on the directed weighted graph according to the PageRank webpage ranking algorithm to obtain the node weights of each of the first nodes and each of the second nodes to be selected in the directed weighted graph; According to the weights of the nodes, at least one target node is determined from the first nodes and the nodes to be selected, so as to determine the customer corresponding to the target node as a potential customer.
8. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 7, characterized in that: The determining, according to the weights of the nodes, at least one target node from the first nodes and the nodes to be selected, so as to determine the customer corresponding to the target node as a potential customer, includes: Sort the first nodes and the nodes to be selected according to their node weights, and obtain rankings of the first nodes and the nodes to be selected; From each of the first nodes and each of the nodes to be selected, each of the target nodes ranked before a preset ranking is obtained, so as to determine each of the customers corresponding to each of the target nodes as a potential customer.
9. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 1, characterized in that: The attribute of the connection path is the transaction summary information between the two customers corresponding to the two connected nodes.
10. The method for screening potential corporate customers of a bank based on an industrial chain map according to claim 9, characterized in that: The direction of the connection path is determined according to the upstream and downstream relationship between two clients corresponding to the two nodes.