A method and system for processing high-net-worth customer information based on relationship network

Abstract

The present invention relates to a method and system for processing high-net-worth customer information based on a relationship network. The method comprises establishing a first community network set based on a network model and a second community network set based on a knowledge graph, and combining the first community network set and the second community network set to obtain a network scenario determination result. The method then performs community discovery on the capital transaction network of high-net-worth customers, performs scenario discovery and determination through social relationships, and performs customized information processing within high-net-worth customers. This method breaks away from conventional supervised machine learning methods, starting from the high-net-worth customer transaction network and performing community discovery based on network dimensional features. This method not only reduces learning costs but also enables more accurate information processing, making it particularly suitable for meeting customer acquisition and product marketing needs.

Description

Technical Field

[0001] The present invention relates to the technical field of financial information data processing, and in particular to a method and system for processing high-net-worth customer information based on a relationship network. Background Art

[0002] High-net-worth customer acquisition and product marketing are key banking businesses. Existing technologies typically employ supervised machine learning for these tasks, using model training, evaluation, and deployment based on detailed historical customer data and a list of high-net-worth customers. However, these conventional models and approaches are no longer sufficient to meet banks' high-net-worth customer marketing needs amidst an increasingly complex economic landscape. In particular, feature engineering for high-net-worth customer marketing models primarily relies on granular static and dynamic features, such as basic customer information, transactional, or product characteristics. Consequently, existing technical models are built from the perspective of the entire customer base, creating a complex and challenging model. This is particularly challenging given the large number of bank customers, resulting in high data computation costs and unsatisfactory marketing results.

[0003] At the same time, the high-net-worth customer marketing model in the existing technology delivers lists of independent individuals, and independent marketing is carried out for individual customers during marketing execution. It does not take into account the transaction network relationship between high-net-worth customers, nor does it take into account the social relationship between high-net-worth customers, and cannot give full play to the role of data. Summary of the Invention

[0004] In order to address the shortcomings of the existing technology, the present invention proposes a high-net-worth customer information processing method and system based on relationship networks, which conducts community discovery for the financial transaction network of high-net-worth customers, discovers and determines scenarios through social relationships, and performs customized information processing within high-net-worth customers. It breaks the conventional supervised machine learning method and starts from the high-net-worth customer transaction network. It conducts community discovery based on network dimensional characteristics. It not only has low learning costs, but also can process information more accurately, which is particularly suitable for meeting customer acquisition and product marketing needs.

[0005] To achieve the above objectives, the technical solutions adopted by the present invention include:

[0006] A method for processing high-net-worth customer information based on a relationship network, characterized by comprising:

[0007] S1. Obtaining original transaction data of a target customer within a preset period, wherein the original transaction data includes at least transaction object information;

[0008] S2. Clean the original transaction data based on the transaction object information, remove the transaction data corresponding to the public exchange, and obtain the target data;

[0009] S3. Establish feature engineering that includes network dimension features and customer basic information dimension features. The network dimension features include network average degree, network density, network transmission rate, network Wiener coefficient, network radius, and network user quality. The customer basic information dimension features include customer asset balance, customer background information, and customer transaction flow data.

[0010] S4. Build a network model based on the target data and feature engineering, and obtain a first community network set in the network model through network clustering identification;

[0011] S5. Determine the relationship between the target customer and other customer entities based on the target data and establish a knowledge graph of the corresponding relationships, and perform a community discovery operation on the knowledge graph to obtain a second community network set;

[0012] S6. Compare the first community network set with the second community network set, extract matching community networks and generate a third community network set;

[0013] S7. Output a network scenario determination result of the target customer based on the third community network set.

[0014] Furthermore, the original transaction data also includes transaction date, transaction amount and transaction type; the preset period is any one of quarterly, semi-annual or annual.

[0015] Furthermore, the step S2 further includes:

[0016] The target data is obtained by performing data quality analysis on the private transaction data retained after removing the transaction data corresponding to the public transaction.

[0017] Furthermore, the step S4 further includes:

[0018] Use Python networkX to build a network model. In the constructed network model, vertices represent target customers and customer entities, edges represent the capital transaction relationship between target customers and customer entities, and the direction of the edge represents the flow of funds.

[0019] The k-means algorithm is used for network clustering. The target customer transaction network that shows a closed-loop transaction pattern is classified as a community network, and the part that shows a concentrated and divergent pattern is classified as an intermediary network.

[0020] Integrate all community networks to obtain the first community network collection.

[0021] Furthermore, the step S5 further includes:

[0022] Use Neo4j to build a knowledge graph;

[0023] Use the GraphX ​​module in Spark to apply the triangular relationship algorithm to perform community discovery operations on the knowledge graph.

[0024] Furthermore, the step S7 further includes:

[0025] Acquire social relationship data of the target customer, wherein the social relationship data includes the type of social relationship between the target customer and the customer entity, wherein the type of social relationship includes relatives, spouses, colleagues, and friends;

[0026] Use the social relationship data to match the third community network set, and generate the network scenario of each community network according to the preset social relationship type combination;

[0027] Integrate the network scenarios of each community network to output the network scenario determination results of the target customers.

[0028] Furthermore, the step S7 further includes:

[0029] Assign relationship weights to social relationship types, and generate the network scenarios of each community network based on the preset social relationship type combinations and relationship weights.

[0030] The present invention also relates to a high-net-worth customer information processing system based on a relationship network, characterized in that it includes:

[0031] Data preprocessing module, used to clean raw transaction data and obtain target data;

[0032] Feature engineering establishment module, used to establish feature engineering including network dimension features and customer basic information dimension features;

[0033] A network model generation module is used to construct a network model based on target data and feature engineering, and obtain a first community network set in the network model through network clustering identification;

[0034] A knowledge graph generation module is used to determine the relationship between the target customer and other customer entities based on the target data and establish a knowledge graph of the corresponding relationship, and perform a community discovery operation on the knowledge graph to obtain a second community network set;

[0035] The network scenario determination module is configured to compare the first community network set with the second community network set, extract matching community networks to generate a third community network set, and output a network scenario determination result of the target customer based on the third community network set.

[0036] The present invention also relates to a computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and the computer program implements the above method when executed by a processor.

[0037] The present invention also relates to an electronic device, characterized in that it comprises a processor and a memory;

[0038] The memory is used to store original transaction data and social relationship data;

[0039] The processor is used to execute the above method by calling original transaction data and social relationship data.

[0040] The beneficial effects of the present invention are:

[0041] The high-net-worth customer information processing method and system based on the relationship network described in the present invention are used to perform community discovery on the capital transaction network of high-net-worth customers, perform scenario discovery and judgment through social relationships, and perform customized information processing within high-net-worth customers. This breaks the conventional supervised machine learning method and starts from the high-net-worth customer transaction network. Based on the network dimension characteristics, community discovery is performed. Not only is the learning cost low, but information can also be processed more accurately, which is particularly suitable for meeting customer acquisition and product marketing needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a flow chart of the method for processing high-net-worth customer information based on a relationship network according to the present invention.

[0043] Figure 2 This is a schematic diagram of the structure of the high-net-worth customer information processing system based on the relationship network of the present invention. DETAILED DESCRIPTION

[0044] In order to more clearly understand the content of the present invention, it will be described in detail with reference to the accompanying drawings and embodiments.

[0045] The first aspect of the present invention relates to a process such as Figure 1 The high-net-worth customer information processing method based on the relationship network shown specifically includes:

[0046] S1. Obtaining raw transaction data of target customers within a preset period (e.g., quarterly, semi-annually, or annually), wherein the raw transaction data includes at least transaction object information and preferably also includes transaction date, transaction amount, and transaction type;

[0047] S2. Clean the original transaction data based on the transaction object information, remove the transaction data corresponding to the public transaction, and perform data quality analysis on the retained private transaction data to obtain the target data;

[0048] S3. Establish feature engineering that includes network dimension features and customer basic information dimension features. The network dimension features include network average degree, network density, network transmission rate, network Wiener coefficient, network radius, and network user quality. The customer basic information dimension features include customer asset balance, customer background information, and customer transaction flow data. Among them, the network average degree (the average degree of customers in the transaction network) can measure the inflow and outflow of each customer, the network density (the proportion of customer relationships in the transaction network) indicates the complexity of customer relationships, the network transmission rate (the number of communities in the transaction network / the maximum number of network communities) indicates the proportion of customers forming communities, the network Wiener coefficient (the average of the sum of the shortest paths between any two points in the transaction network) indicates the scalability of the network, and the network radius (the maximum value of all shortest paths in the transaction network) indicates the overall size of the graph. The network user quality can be measured by the proportion of private banking customers on the network to measure the overall quality of the network.

[0049] S4. Use Python networkX to build a network model based on the target data and feature engineering, and use the k-means algorithm to perform network clustering to obtain the first community network set in the network model. Specifically, the portion of the target customer transaction network that exhibits a closed-loop transaction pattern is classified as a community network, and the portion that exhibits a concentrated and divergent pattern is classified as an intermediary network.

[0050] S5. Determine the relationship between the target customer and other customer entities based on the target data and use Neo4j to build a knowledge graph of the corresponding relationships. Use the GraphX ​​module in Spark to apply a triangular relationship algorithm to perform a community discovery operation on the knowledge graph to obtain a second community network set. In particular, for two different customer entities that are both connected to the target customer, if the two different customer entities also have a connection, the target customer and the two different customer entities are identified as a community network. Depending on actual computing needs, the Louvain algorithm, label propagation algorithm, etc. can also be used to replace the triangular relationship algorithm to perform the community discovery operation.

[0051] S6. Compare the first community network set with the second community network set, extract matching community networks and generate a third community network set;

[0052] S7. Outputting a network scenario determination result for the target customer based on the third community network set, specifically including obtaining the target customer's social relationship data. The social relationship data includes the types of social relationships between the target customer and the customer entity. Sources may include debit and credit card contacts, trusted transferees, and contacts with the same contact information or address. The social relationship types include relatives, spouses, colleagues, and friends. Relationship weights are assigned to social relationship types. For example, the relationship weights for relatives and spouses may be set higher than those for colleagues and friends to indicate a closer connection. The social relationship data is used to match the third community network set, and the network scenario to which each community network belongs is generated based on a preset combination of social relationship types and relationship weights. For example, social relationships primarily consisting of relatives or spouses are determined to be family community network scenarios, while social relationships primarily consisting of colleagues or friends are determined to be intermediary community network scenarios. The network scenarios of each community network are integrated to output a network scenario determination result for the target customer. For family community network scenarios, family trust products or customer acquisition marketing may be recommended. For intermediary community networks, high-end financial management products or customer acquisition marketing may be recommended. At the same time, marketing suggestions and lists can be output from network scenarios and sent to downstream systems, so that the network model and its scenarios can be updated in a timely manner based on the execution effect of the list.

[0053] Another aspect of the present invention relates to a structure such as Figure 2 The high-net-worth customer information processing system based on the relationship network shown specifically includes:

[0054] Data preprocessing module, used to clean raw transaction data and obtain target data;

[0055] Feature engineering establishment module, used to establish feature engineering including network dimension features and customer basic information dimension features;

[0056] A network model generation module is used to construct a network model based on target data and feature engineering, and obtain a first community network set in the network model through network clustering identification;

[0057] A knowledge graph generation module is used to determine the relationship between the target customer and other customer entities based on the target data and establish a knowledge graph of the corresponding relationship, and perform a community discovery operation on the knowledge graph to obtain a second community network set;

[0058] The network scenario determination module is configured to compare the first community network set with the second community network set, extract matching community networks to generate a third community network set, and output a network scenario determination result of the target customer based on the third community network set.

[0059] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention are intended to be covered by the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for processing high-net-worth customer information based on a relationship network, characterized in that: include: S1. Obtaining original transaction data of a target customer within a preset period, wherein the original transaction data includes at least transaction object information; S2. Clean the original transaction data based on the transaction object information, remove the transaction data corresponding to the public exchange, and obtain the target data; S3. Establish feature engineering that includes network dimension features and customer basic information dimension features. The network dimension features include network average degree, network density, network transmission rate, network Wiener coefficient, network radius, and network user quality. The customer basic information dimension features include customer asset balance, customer background information, and customer transaction flow data. S4. Build a network model based on the target data and feature engineering, and obtain a first community network set in the network model through network clustering identification; S5. Determine the relationship between the target customer and other customer entities based on the target data and establish a knowledge graph of the corresponding relationships, and perform a community discovery operation on the knowledge graph to obtain a second community network set; S6. Compare the first community network set with the second community network set, extract matching community networks and generate a third community network set; S7. Outputting a network scenario determination result of the target customer based on the third community network set; Wherein, the step S4 further includes: Use Python networkX to build a network model. In the constructed network model, vertices represent target customers and customer entities, edges represent the capital transaction relationship between target customers and customer entities, and the direction of the edge represents the flow of funds. The k-means algorithm is used for network clustering. The target customer transaction network that shows a closed-loop transaction pattern is classified as a community network, and the part that shows a concentrated and divergent pattern is classified as an intermediary network. Integrate all community networks to obtain the first community network set; The step S5 further includes: Use Neo4j to build a knowledge graph; Use the GraphX ​​module in Spark to apply the triangular relationship algorithm to perform community discovery operations on the knowledge graph; The step S7 further comprises: Acquire social relationship data of the target customer, wherein the social relationship data includes the type of social relationship between the target customer and the customer entity, wherein the type of social relationship includes relatives, spouses, colleagues, and friends; Use the social relationship data to match the third community network set, and generate the network scenario of each community network according to the preset social relationship type combination; Integrate the network scenarios of each community network to output the network scenario determination results of the target customers.

2. The method according to claim 1, wherein The original transaction data also includes transaction date, transaction amount and transaction type; the preset period is any one of quarterly, semi-annual or annual.

3. The method according to claim 1, wherein The step S2 further includes: The target data is obtained by performing data quality analysis on the private transaction data retained after removing the transaction data corresponding to the public transaction.

4. The method according to claim 1, wherein The step S7 further comprises: Assign relationship weights to social relationship types, and generate the network scenarios of each community network based on the preset social relationship type combinations and relationship weights.

5. A high net worth customer information processing system based on a relationship network, applying the method according to claim 1, characterized in that: include: Data preprocessing module, used to clean raw transaction data and obtain target data; Feature engineering establishment module, used to establish feature engineering including network dimension features and customer basic information dimension features; A network model generation module is used to construct a network model based on target data and feature engineering, and obtain a first community network set in the network model through network clustering identification; A knowledge graph generation module is used to determine the relationship between the target customer and other customer entities based on the target data and establish a knowledge graph of the corresponding relationship, and perform a community discovery operation on the knowledge graph to obtain a second community network set; The network scenario determination module is configured to compare the first community network set with the second community network set, extract matching community networks to generate a third community network set, and output a network scenario determination result of the target customer based on the third community network set.

6. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which implements the method according to any one of claims 1 to 4 when executed by a processor.

7. An electronic device, characterized in that: including processor and memory; The memory is used to store original transaction data and social relationship data; The processor is configured to execute the method according to any one of claims 1 to 4 by calling original transaction data and social relationship data.

Images