A social graph construction method and device based on differential privacy and related products

Abstract

The application provides a social graph construction method and device based on differential privacy and related products, and relates to the technical field of data processing. The social graph construction method based on differential privacy acquires a target business requirement, and disassembles the target business requirement into a target data requirement set; conducts compliance risk assessment on the data requirements in the target data requirement set, and generates a data collection and processing strategy; differentially privatizes the original data by using a differential privacy strategy to obtain a target data set; constructs a social network graph based on the target data set; and finally generates and executes a business decision rule based on the social network graph. The application solves the contradiction between business data value mining and compliance risk prevention and control, realizes data closed-loop processing from a business target, with compliance embedded and privacy protection, and directly enables precise marketing and risk control and other core businesses.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus and related products for constructing a social graph based on differential privacy. Background Technology

[0002] In an increasingly competitive financial services industry, banks are extending their customer engagement beyond traditional financial services to multiple dimensions of customer operations and lives, and further into customer social networks. Currently, the industry is generally attempting to integrate internal bank data with external data sources to conduct in-depth analysis of individual customer profiles and their social network profiles. The aim is to leverage social networks to achieve efficient viral marketing or improve the accuracy and coverage of customer and transaction risk identification.

[0003] However, customer social network data often involves sensitive personal privacy, and banks must ensure the legal and compliant acquisition and use of this data while strictly adhering to relevant laws and regulations on data security and personal information protection. In practice, some bank mobile applications have been subject to regulatory notices or penalties for excessive collection of customer data, highlighting the shortcomings in the banking industry's data compliance practices. Meanwhile, as regulatory requirements for customer risk management become increasingly stringent, remote banking and online business transactions place higher demands on customer identity authentication and security control. This requires banks to expand their ability to construct personal profiles of customers beyond those provided by customers themselves or by authoritative institutions, and customer social networks are a key information dimension for creating such profiles.

[0004] Currently, the following technical shortcomings exist in the process of building and applying customer social networks: (1) Data collection is out of sync with business objectives, posing significant compliance risks: Existing solutions often adopt broad and general data collection strategies, lacking precise planning oriented towards specific business objectives. Due to the failure to clearly define the minimum necessary data set required to achieve business objectives, the scope of data collection is too large, which not only increases data processing costs but also creates compliance risks associated with excessive data collection; (2) Balancing privacy protection and model accuracy is difficult: Traditional technical approaches often face a dilemma. On the one hand, excessive use of data anonymization or desensitization to meet compliance requirements can easily lead to sparse social network relationships, loss of key features, and a significant decrease in model accuracy. On the other hand, directly processing raw sensitive data in pursuit of a high-precision model carries an extremely high risk of data misuse. There is a lack of technical means to effectively extract high-quality relationship features without accessing the raw data. (3) The general model has low fit with banking business scenarios: the existing general social network analysis model has failed to deeply integrate into the unique business logic of the banking industry (such as fund transfer, guarantee link, product holding relationship, etc.), making it difficult to define and build accurate and comprehensive social relationships that truly meet the needs of banking risk control, marketing and other scenarios, resulting in insufficient businessization of the model.

[0005] Therefore, achieving an effective balance between compliant data processing and the construction of high-precision social networks has become a technical challenge that urgently needs to be overcome. Summary of the Invention

[0006] In view of the above problems, this application is proposed to provide a method, apparatus, and related product for constructing a social graph based on differential privacy that overcomes or at least partially solves the above problems. The technical solution is as follows: Firstly, a method for constructing a social graph based on differential privacy is provided, the method comprising: Obtain the target business requirements, and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement; Based on preset indicators, a risk assessment is performed on the data requirements of the target data requirement set to generate a data collection and processing strategy. Based on the data acquisition and processing strategy, raw data is acquired, and the raw data is then processed for privacy using a differential privacy strategy to obtain the target dataset. Construct a social network graph based on the target dataset; Business decision rules are constructed based on the social network graph, and target business operations are executed.

[0007] In one possible implementation, obtaining the target business requirements and generating a target data requirement set based on the target business requirements includes: Based on the mind chain guidance model, the target business requirement is broken down into at least one sub-target; For each of the sub-objectives, at least one associated business metric shall be identified; Based on each of the aforementioned business metrics, data requirement items for constructing a social network graph are mapped to form the target data requirement set.

[0008] In one possible implementation, the step of conducting a risk assessment of the data requirements of the target data requirement set based on preset indicators and generating a data collection and processing strategy includes: Obtain the authorization terms information corresponding to the data requirements of the target data requirement set; Based on the authorization terms information, a mapping matrix is ​​established between the authorization terms information and data usage, and the mapping matrix is ​​used to represent the data usage permitted by each authorization term; Based on the mapping matrix and in accordance with the principle of minimum necessity, each data requirement in the target data requirement set is evaluated for purpose relevance, means minimumity, and scope applicability to obtain the risk assessment result corresponding to each data requirement. Based on the risk assessment results of all data requirements, the data collection and processing strategy is generated.

[0009] In one possible implementation, the privacy processing of the original data using a differential privacy strategy includes: Identify the data type of the raw data, which includes numerical data, relational data, and categorical and location data; Add Laplace noise to numerical data; Add noise to the edge weights in relational data, or perform random reconnection of the graph structure; An exponential mechanism is used to fuzzify categorical and location-based data.

[0010] In one possible implementation, constructing the social network graph based on the target dataset includes: Based on the target dataset, the constituent elements required to construct the social network graph are identified. The constituent elements include a set of nodes and a set of relation edges. Each node in the set of nodes corresponds to an entity object, and the node contains the unique identification information and attribute information of the entity object. Based on the data relationships contained in the target dataset, define the relationship edges between the nodes in the node set, and determine the type and strength of each relationship edge; wherein, the type of relationship edge includes at least one of business association relationship, behavior association relationship, and attribute similarity relationship; The social network graph is constructed based on the set of nodes, the set of relation edges, and the type and strength of each relation edge.

[0011] In one possible implementation, the step of constructing business decision rules based on the social network graph and executing target business operations includes: Based on the social network graph, target nodes with specific business attributes are identified; Based on the business attributes of the target node, generate the appropriate business decision rules; Based on the aforementioned business decision rules, the corresponding target business operation is triggered and executed.

[0012] In one possible implementation, the method further includes: Attribution analysis or A / B testing is performed on the execution effect of the target business operation to obtain the effect evaluation results; Based on the effect evaluation results, the target data requirement set, the data collection and processing strategy, or the social network graph are iteratively optimized.

[0013] Secondly, a social graph construction device based on differential privacy is provided, the device comprising: A requirement decomposition unit is used to obtain target business requirements and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement. The risk assessment unit is used to assess the risk of data requirements for the target data requirement set based on preset indicators and generate data collection and processing strategies. The privacy processing unit is used to acquire raw data based on the data acquisition and processing strategy, and to perform privacy processing on the raw data using a differential privacy strategy to obtain the target dataset. The graph construction unit is used to construct a social network graph based on the target dataset; The application execution unit is used to construct business decision rules based on the social network graph and execute target business operations.

[0014] Thirdly, an electronic device is provided, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the differential privacy-based social graph construction method described in any of the preceding claims.

[0015] Fourthly, a storage medium is provided that stores a computer program, wherein the computer program is configured to execute the differential privacy-based social graph construction method described above at runtime.

[0016] By employing the above technical solutions, the differential privacy-based social graph construction method, apparatus, and related products provided in this application embodiment acquire target business requirements and decompose them into a target data requirement set; conduct compliance risk assessment on the data requirements in the target data requirement set to generate data collection and processing strategies; use differential privacy strategies to perform differentiated privacy processing on the original data to obtain a target dataset; construct a social network graph based on the target dataset; and finally generate and execute business decision rules based on the social network graph. This application embodiment resolves the contradiction between business data value mining and compliance risk prevention and control, and realizes closed-loop data processing that starts from business objectives, incorporates compliance, and protects privacy, directly empowering core businesses such as precision marketing and risk management. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0018] Figure 1 A flowchart of a social graph construction method based on differential privacy provided in an embodiment of this application is shown; Figure 2 An architecture diagram of the social graph construction method based on differential privacy provided in a specific embodiment of this application is shown; Figure 3 A flowchart illustrating a social graph construction method based on differential privacy provided in a specific embodiment of this application is shown; Figure 4 This paper illustrates a flowchart of the target business requirements breakdown for a social graph construction method based on differential privacy provided in a specific embodiment of this application. Figure 5 The flowchart illustrating the requirement data item completion process of the social graph construction method based on differential privacy provided in a specific embodiment of this application is shown. Figure 6 This diagram illustrates the structure of a social graph construction apparatus based on differential privacy provided in an embodiment of this application. Figure 7 A structural diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0019] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0021] The inventors discovered through analysis that in cloud infrastructure architectures, computing, storage, and network resources are provided in a pooled manner, with upper-layer virtualization resources and container resources dynamically scheduled and created based on these underlying resource pools. Although there are close carrying and dependency relationships between physical servers, virtual machines, and container applications, their management data and operation and maintenance interfaces are scattered across independent configuration management databases, cloud management consoles, and container management consoles. The inconsistent data models and fragmented association identifiers across platforms result in the complete resource topology from the physical layer to the application layer being fragmented and stored in multiple heterogeneous systems.

[0022] When a single physical server fails, its impact propagates upwards along the resource topology chain, from the physical server to virtual machines and container applications. However, because this resource topology is broken at the management level, operations personnel cannot directly and automatically obtain the complete list of affected virtual machines and all the container applications they host, nor can they automatically orchestrate verification tasks covering the entire chain. This forces post-failure verification to rely on manual cross-referencing and combination operations across multiple systems, resulting in low efficiency and a high risk of errors.

[0023] Therefore, the core issue is: how to automatically collect, associate, and integrate the resource status and relationships scattered across various independent management platforms in order to reconstruct and maintain a unified and accurate cross-level resource logical topology, and based on this topology, realize the automatic determination of the scope of fault impact and the intelligent orchestration of verification tasks. This points to the direction of technological breakthroughs for subsequent solutions.

[0024] To address the aforementioned technical problems, embodiments of this application provide a social graph construction method based on differential privacy, such as... Figure 1 As shown, the differential privacy-based social graph construction method may include the following steps S101 to S105: Step S101: Obtain the target business requirements and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement.

[0025] In one possible implementation, the aforementioned target business requirements can be strategic business intentions described in natural language, such as improving the performance of a specific business line, controlling a certain type of risk, or optimizing customer experience.

[0026] In one possible implementation, the process of generating a target data requirement set based on target business needs can be automated by using an artificial intelligence model that can understand the business logic and intent described in natural language.

[0027] In one example, if the target business requirement is to enhance the risk control capabilities of retail credit business, the artificial intelligence model can break it down into sub-objectives such as identifying potential fraudulent applications and monitoring the deterioration of existing customer credit. For the sub-objective of identifying potential fraudulent applications, business indicators such as the fraud case identification rate and the correlation with gang fraud can be derived, thereby mapping out data items such as application device information, application geographical location sequence, and abnormal fund transaction records in the customer's associated network that need to be collected.

[0028] In another example, if the target business involves increasing the asset size of wealth management clients, it can be broken down into sub-objectives such as identifying high-potential clients and promoting cross-holding of products; the corresponding data requirements may include client asset size estimation signals from other banks, client historical product purchase preferences, and client social influence centrality indicators within the enterprise.

[0029] In one possible implementation, each data requirement in the aforementioned target data requirement set may include business meaning, expected format, and correlation with upstream business sub-targets and indicators. This embodiment does not impose any restrictions on this.

[0030] Step S102: Based on preset indicators, conduct a risk assessment of the data requirements of the target data requirement set and generate a data collection and processing strategy.

[0031] In one possible implementation, the aforementioned risk assessment can be understood as a process of systematically analyzing and evaluating the compliance, security, and necessity of data requirements based on laws, regulations, regulatory requirements, and the bank's internal compliance policies.

[0032] In one possible implementation, the aforementioned data collection and processing strategy can be understood as a set of normative instructions or rules that guide subsequent data acquisition and privacy processing operations. Its core objective is to ensure that data activities meet business needs while complying with compliance requirements.

[0033] Step S103: Based on the data acquisition and processing strategy, acquire the raw data, and use the differential privacy strategy to perform privacy processing on the raw data to obtain the target dataset.

[0034] In one possible implementation, the aforementioned data collection and processing strategy can be understood as the compliance requirements, processing methods, and technical parameters that need to be followed when processing raw data.

[0035] In another possible implementation, this step can be understood as a process of protectively transforming the original data that meets business needs based on the compliance and technical requirements determined in the previous steps. Without directly exposing the specific values ​​of the original data, a target dataset that can be safely used for subsequent analysis is generated through controlled noise injection and other methods. This dataset contains the statistical characteristics and relational patterns of the original data and can effectively prevent individual information from being reverse-inferred.

[0036] Step S104: Construct a social network graph based on the target dataset.

[0037] Step S105: Construct business decision rules based on the social network graph and execute the target business operation.

[0038] In one possible implementation, the aforementioned business decision-making rules can be understood as identifying patterns, groups, or paths that meet preset business conditions by quantitatively or qualitatively analyzing the node characteristics, relationship strength, and network topology of the social network graph, and forming executable strategies or rules accordingly.

[0039] In one possible implementation, executing the target business operation can be understood as transforming the completed business decision rules into specific system instructions or manual operation processes, and running them in a designated business system or channel, thereby achieving business objectives such as managing a specific customer group, promoting a certain service, or intervening in potential risks.

[0040] In one possible implementation, the execution of the aforementioned business decision rules can be automated. For example, after the system identifies a subgraph that conforms to the risk diffusion pattern, it automatically generates and triggers a risk warning rule, which can lead to the system automatically implementing temporary control over the transaction permissions of the relevant accounts.

[0041] In one possible implementation, the execution of the aforementioned business decision rules can also be semi-automatic or decision-assisted. For example, the system identifies customer groups that may have potential cross-selling opportunities and generates product recommendation rules and a customer list. This list and rules are pushed to the workbench of business personnel, who then review them, manually select them, and execute batch marketing operations.

[0042] This embodiment acquires target business requirements and breaks them down into a target data requirement set; it conducts a compliance risk assessment on the data requirements in the target data requirement set and generates a data collection and processing strategy; it uses a differential privacy strategy to perform differentiated privacy processing on the original data to obtain the target dataset; it constructs a social network graph based on the target dataset; and finally, it generates and executes business decision rules based on the social network graph. Furthermore, it resolves the contradiction between business data value mining and compliance risk prevention and control, and realizes a closed-loop data processing that starts from business objectives, incorporates compliance, and protects privacy, directly empowering core businesses such as precision marketing and risk management.

[0043] This application embodiment provides a possible implementation method. The above step S101, which obtains the target business requirements and generates a target data requirement set based on the target business requirements, may specifically include the following steps A1 to A3: Step A1: Based on the mind chain guidance model, break down the target business requirement into at least one sub-target; Step A2: For each sub-objective, identify at least one associated business metric; Step A3: Based on each business metric, map out the data requirement items for building the social network graph, forming the target data requirement set.

[0044] This embodiment provides a method for transforming abstract target business requirements into a specific, executable set of data requirements. By constructing an automated and standardized mapping link from business objectives to data requirements, it effectively solves the problem of disconnect between business intent and underlying data support.

[0045] This application embodiment provides a possible implementation method. The above step S102 performs a risk assessment on the data requirements of the target data requirement set based on preset indicators and generates a data collection and processing strategy, which may specifically include the following steps B1 to B4: Step B1: Obtain the authorization terms information corresponding to the data requirements of the target data requirement set; Step B2: Based on the authorization terms information, establish a mapping matrix between the authorization terms information and the data usage. The mapping matrix is ​​used to represent the data usage permitted by each authorization term. Step B3: Based on the mapping relationship matrix and in accordance with the principle of minimum necessity, conduct a purpose relevance assessment, a means minimum assessment, and a scope applicability assessment for each data requirement in the target data requirement set to obtain the risk assessment result corresponding to each data requirement. Step B4: Generate a data collection and processing strategy based on the risk assessment results of all data requirements.

[0046] This embodiment constructs a structured risk assessment and strategy generation mechanism, transforming abstract legal and regulatory requirements and privacy compliance requirements into specific, executable, and auditable data operation instructions. This establishes a systematic bridge between business-driven data needs and a strict data protection compliance framework, forming a management closed loop from rule parsing and risk quantification to strategy generation, ensuring the legality, compliance, and risk controllability of data activities from the source.

[0047] This application embodiment provides a possible implementation method. The above step S103 uses a differential privacy strategy to perform privacy processing on the original data, which may specifically include the following steps C1 to C4: Step C1: Identify the data type of the original data. Data types include numerical data, relational data, and categorical and location data. Step C2: Add Laplace noise to the numerical data; Step C3: Add noise to the edge weights in the relational data, or perform random reconnection of the graph structure; Step C4: Use an exponential mechanism to fuzzify the categorical and location-based data.

[0048] This embodiment provides a complete technical path for systematically implementing differential privacy protection. By accurately identifying and classifying data of different natures and applying appropriate and differentiated privacy protection mechanisms, it maximizes the preservation of statistical features and relational structures in the dataset that are valuable for subsequent construction of social network graphs, while strictly protecting sensitive information in the original data.

[0049] This application embodiment provides a possible implementation method. The above step S104, which constructs a social network graph based on the target dataset, may specifically include the following steps D1 to D3: Step D1: Based on the target dataset, identify the constituent elements required to construct the social network graph. The constituent elements include a set of nodes and a set of relation edges. Each node in the node set corresponds to an entity object, and the node contains the unique identification information and attribute information of the entity object. Step D2: Based on the data relationships contained in the target dataset, define the relationship edges between nodes in the node set, and determine the type and strength of each relationship edge; wherein, the type of relationship edge includes at least one of business association relationship, behavior association relationship, and attribute similarity relationship; Step D3: Construct a social network graph based on the set of nodes, the set of relation edges, and the type and strength of each relation edge.

[0050] This embodiment organizes a set of nodes carrying identifiers and attributes, and a set of relation edges with type and strength, into an organic whole, namely a social network graph, based on a graph theory model. This graph clearly presents the multi-dimensional network of relationships between entities with a visual and computable data structure. The graph constructed in this way carries a structured knowledge representation with rich business semantics and relation weights, which can be directly applied to various high-level business scenarios such as influence customer mining, community discovery, and risk transmission path analysis, providing direct and powerful network insight support for decision-making such as precision marketing and risk prevention.

[0051] This application provides a possible implementation method. Step S105 above constructs business decision rules based on the social network graph and executes the target business operation, which may specifically include the following steps E1 to E3: Step E1: Identify target nodes with specific business attributes based on the social network graph; Step E2: Generate suitable business decision rules based on the business attributes of the target node; Step E3: Based on the business decision rules, trigger and execute the corresponding target business operation.

[0052] This embodiment transforms the completed social network graph—a data asset—into a decision-making basis that can directly drive business actions, thus realizing a value loop from relationship insight to business execution.

[0053] This application provides a possible implementation method. The above-described social graph construction method based on differential privacy may further include the following steps F1 to F2: Step F1: Perform attribution analysis or A / B testing on the execution effect of the target business operation to obtain the effect evaluation results; Step F2 involves iteratively optimizing the target data requirement set, data collection and processing strategy, or social network graph based on the effect evaluation results.

[0054] This application's embodiments introduce an effect evaluation and closed-loop optimization mechanism, systematically transforming business effect feedback into a driving force for the continuous evolution of data planning, compliance strategies, graph models, and business rules. This ensures that the efficiency of data-to-business value conversion is continuously improved in a dynamic environment, achieving a leap from static construction to dynamic self-optimization.

[0055] The above introduces Figure 1 The embodiments shown have various implementation methods for each stage. The following will further explain the social graph construction method based on differential privacy of this application through specific embodiments.

[0056] like Figure 2As shown, this specific embodiment is a three-tiered, multi-module-linked intelligent risk control and business empowerment architecture, including a basic layer, a system capability layer, and an application layer.

[0057] The foundational layer provides underlying support, including model capabilities such as large-scale pre-trained models, natural language processing, and knowledge graphs; data capabilities such as data governance, feature engineering, and tagging systems; heterogeneous big data computing capabilities such as distributed computing, Hadoop (a distributed file system and distributed computing framework), and Elasticsearch (a distributed search and analysis engine); and provides the computing power, data, and algorithm foundation for upper-layer applications.

[0058] System Capability Layer: This is the core hub, divided into four major sections: intelligent risk control, intelligent marketing, intelligent compliance, and intelligent anti-fraud. Intelligent risk control: covers the entire process of pre-loan, during-loan, and post-loan, and also has capabilities such as account security; Intelligent marketing: Achieving customer acquisition and conversion through user profiling, precision marketing, and intelligent recommendations; Intelligent compliance: Achieving compliance with laws and regulations, financial compliance, data security, and privacy protection; Intelligent anti-fraud: It prevents fraud risks through multi-dimensional risk control, data verification, and intelligent verification.

[0059] Application layer: Transforming system capabilities into specific business value, including key business scenarios such as CRM (Customer Relationship Management) management and external credit granting; innovative products such as supply chain finance, open banking, and enterprise-level settlement; regulatory applications such as enterprise credit management, regulatory reporting, and intelligent early warning; and achieving deep integration of risk control and business.

[0060] like Figure 3 As shown, the implementation process of this specific embodiment is as follows: I. Target Business Requirements Decomposition and Target Data Requirements Set Generation Phase 1. Identify target business needs: Conduct efficient viral marketing campaigns for banks based on customer social networks, and control fraud risks; in response to recent high incidence of risk events such as "coupon hunters," enhance the breadth and depth of risk analysis and increase the mining and analysis of customer social networks.

[0061] 2. Based on the mind chain guidance model, the above target business requirements are broken down into two sub-objectives: Sub-objective one is "to build a normalized referral high-quality customer acquisition model", and sub-objective two is "to identify fraudulent groups that exploit loopholes".

[0062] 3. For example Figure 4 As shown, perform the following operations for the sub-target: (1) Sub-target 1 Identify relevant business metrics: K-factor > 1.2, referral conversion rate decrease of 20%, etc. Based on business metrics, the social networking strategy is determined to identify and incentivize influencers; As shown in Table 1, the final data requirements include customer influence value, customer sharing behavior, customer loyalty, and referral probability. Table 1 - Sub-target Data Requirements

[0063] (2) Sub-objective two Identifying relevant business metrics: Fraud gang early detection rate increased by 50%; Based on business metrics, the social network strategy was determined to be the detection of abnormal subgraphs and clustering patterns. The final data requirements for mapping are: device clustering, transaction loop, relationship mutation, etc.

[0064] 4. Integrate all the above data requirements to form the target data requirement set.

[0065] II. Data Acquisition and Processing Strategy Generation Stage 1. Conduct a comprehensive inventory of the bank's existing data assets, record data information, and obtain the information shown in Table 2: Table 2 - Data Asset Information Record Table

[0066] 2. For example Figure 5 As shown, the existing data assets after inventory are compared with the target data requirement set to determine the type of data gap and complete the data item supplementation: Scenario 1 (Completely Missing): The data requirement item "centrality data in social network feature vectors" has no direct data item and cannot be obtained by processing existing data; Scenario 2 (existing and authorized coverage): The data requirements "normal referral closed-loop data", "daily transfer transaction data", "equipment aggregation data", "transaction loop data", "relationship mutation data" and "product co-investment relationship data" all exist, and the authorized scope covers the corresponding business uses; Scenario 3 (Existing but incomplete authorization): The "Shared link click-through rate data" in the data requirement item "Historical referral success rate data" already exists, but the authorization scope only includes marketing purposes and does not cover extended needs for risk control purposes.

[0067] 3. Develop supplementary plans for different types of gaps. For scenario one: assess the feasibility of replacing the proxy variable. For example, if "customer social relationship interaction frequency data + percentage of successful referrals" is determined as the proxy variable, the assessment result is that the loss rate is ≤10%. For scenario three: Initiate the expanded authorization assessment process, add authorization terms for risk control purposes to the privacy policy, and simultaneously complete the user notification process; Integrate data asset inventory results, data gap analysis, supplementation plans, and authorization requirements to generate a data authorization requirement report, which clarifies the acquisition path, authorization status, and supplementation deadline for each data requirement item.

[0068] 4. Obtain data authorization terms and data authorization requirement reports corresponding to the target data requirement set, including data authorization terms for marketing and risk control purposes in the bank's privacy policy and user agreement.

[0069] 5. Based on the authorization terms, establish a mapping matrix between the authorization terms and data usage: For example, map the "Recommend products or services that you are interested in" term to "Marketing" usage, and the corresponding data requirements are routine referral closed-loop data, data related to identifying influential customers, and data related to identifying vulnerable customers; map the "Protect your account security and prevent fraud risks" term to "Risk Control" usage, and the corresponding data requirements are device aggregation data, transaction loop data, and relationship mutation data.

[0070] 6. Based on the mapping matrix and data authorization requirement report, evaluate each data requirement according to the principle of minimum necessity: (1) Purpose relevance assessment: All data requirements are essential for achieving the sub-goals; (2) Minimalism assessment: There is no alternative data that does not require authorization or is not sensitive, and the corresponding data requirements need to be collected; (3) Scope applicability assessment: The scope of data collection is limited to the bank's existing customers and transaction counterparties. The accuracy and granularity of the data are based on meeting the requirements of indicator calculation and customer identification. For example, only data related to customer referral behavior are collected, and irrelevant social information is not collected.

[0071] 7. Generate data acquisition and processing strategies based on evaluation results. (1) Data collection method: Extract routine referral closed-loop data, daily transfer transaction data, and product co-investment data from the bank's core business system; extract social relationship matrix data from the social relationship building module; and supplement proxy variable data according to the data authorization requirement report. (2) Processing requirements: The collected raw data shall be classified and labeled according to the sensitivity level. For example, customer identification and transaction amount are the most sensitive data and require special privacy protection processing.

[0072] III. Raw Data Privacy Processing and Target Dataset Generation Stage 1. Identify the data type of the raw data (1) Numerical data: transaction amount, transaction frequency, referral rate, click rate, registration rate, number of accounts opened, K factor value, number of transactions in 30 days, transaction amount in 30 days, and referral cost amount; (2) Relational data: customer-to-customer financial trust relationship edges, product co-investment relationship edges, family circle relationship edges, workplace circle relationship edges, and guarantee circle relationship edges; (3) Classification and location data: customer occupation, home address, workplace address, and business type.

[0073] 2. Differential privacy strategies are used to process different types of data. (1) Logarithmic data: Continuous noise is added using the Laplace mechanism. For example, the original transaction amount of 10,000 yuan becomes 10,023 yuan after adding noise (noise value +23), and the original referral rate of 30% becomes 31.2% after adding noise (noise value +1.2%). (2) For relational data: add Laplace noise to the weight of each relational edge. For example, the weight of the financial trust relational edge is 0.8, and after adding noise, it becomes 0.78. At the same time, delete some weak relational edges with a 5% random probability and add irrelevant weak relational edges with a 3% random probability. (3) For classification and location data: use an index mechanism to fuzzify the data, such as fuzzing the precise home address "XX City XX District XX Street XX No." into "XX City XX District Grid A3 Area", and adding Laplace noise to the statistical occupational distribution data.

[0074] 3. Integrate all the privacy-processed data to form the target dataset.

[0075] IV. Social Network Graph Construction Stage 1. Identify the constituent elements of the social network graph based on the target dataset. (1) Node set: includes bank customer nodes and potential customer nodes (trading counterparties, etc.). The unique identifier of each node is the user ID. The attribute information includes social identifier, consumption preference tag, customer risk level, and wealth level. For example, customer node "ID-001", social identifier "user nickname: XX", consumption preference tag "financial preference", customer risk level "low risk", and wealth level "middle class". (2) Set of relation edges: derived from the data relationships in the target dataset, including financial relation edges (transfer, transaction, joint loan, guarantee), behavioral relation edges (shared equipment, overlapping geographical locations, business processing sequence), and other relation edges (joint product ownership, similar risk preferences, similar unit addresses, similar home addresses).

[0076] 2. Define the type and strength of the edges that relate nodes. (1) Types of relationship edges: financial relationship edges, behavioral relationship edges, and attribute similarity relationship edges; (2) Relationship edge strength: determined based on transaction frequency, amount, interaction frequency, etc. For example, if customers “ID-001” and “ID-002” transfer money 5 times a month, the financial relationship edge strength is 0.9; if customers “ID-003” and “ID-004” belong to the same workplace circle and their business processing time highly overlaps, the behavioral relationship edge strength is 0.7.

[0077] 3. Based on the set of nodes, the set of relation edges, and the type and strength of relation edges, construct a social network graph. The graph includes sub-networks such as family circles, colleague circles, guarantee circles, and neighborhood relationship circles. For example, "ID-001", "ID-005", and "ID-006" form a family circle network because their family addresses are similar and they have financial transactions.

[0078] V. Business Decision-Making Rule Construction and Target Business Operation Execution Phase 1. Identifying target nodes with specific business attributes based on social network graphs (1) Identify high-influence customer nodes: Select nodes with degree centrality in the top 10%, eigenvector centrality ≥ 0.8, and historical referral success rate ≥ 40%, such as “ID-001”, “ID-012”, “ID-035”, and label them as “marketing campaign recommender”; (2) Identify high-value potential customer nodes: In the social relationship chain of high-value customers (wealth level "high net worth"), screen out nodes that have not yet opened an account with the bank, such as the social relationship nodes "ID-051" and "ID-052" of high-value customer "ID-050", and label them as "high-value potential customers"; (3) Identify the nodes of the "wool party" gang: Screen the node clusters with device aggregation degree ≥5, abnormal transaction loops, and relationship mutation times ≥3, such as the node clusters composed of "ID-080", "ID-081" and "ID-082", and label them as "suspected wool party gang".

[0079] 2. Generate suitable business decision rules based on the business attributes of the target node. (1) For the “Marketing Activity Referrer” node: Generate viral marketing rules, including providing exclusive invitation codes, referrers can receive points rewards after successfully referring new customers to open an account (1,000 points for each referrer), and new customers can receive an opening gift; (2) For the “high-value potential customer” node: generate targeted customer acquisition rules, including customer managers taking the initiative to make phone calls, providing exclusive financial product recommendations, and enjoying preferential rates when opening an account; (3) For nodes suspected of being "coupon-sharing groups": generate risk control rules, including restricting participation in marketing activities, suspending coupon issuance, and focusing on monitoring transaction behavior.

[0080] 3. Execute target business operations based on business decision rules. (1) Push exclusive invitation codes to the “Marketing Campaign Recommendation Officer” node and synchronize them to the account managers under their management to launch a viral marketing campaign; (2) Assign the task of reaching out to "high-value potential customers" to the account manager and start the targeted customer acquisition process; (3) Restrict marketing activities for nodes suspected of being "wool party gangs" and include them in the high-risk monitoring list.

[0081] VI. Effect Evaluation and Iterative Optimization Phase 1. Evaluate the effectiveness of the target business operations. (1) Attribution analysis: For example, by tracing the new customer account opening process, it was determined that the number of new customers who opened accounts through the "marketing campaign referral officer" was 850, accounting for 34% of the total number of new customers; (2) A / B testing: If customers are divided into two groups, Group A adopts a social network graph-driven marketing approach, while Group B adopts a traditional marketing approach. The test period is 3 months. The results show that the referral cost of Group A is 22% lower than that of Group B, the K factor reaches 1.3, and the early identification rate of fraud gangs is increased by 53%.

[0082] 2. Iterative optimization based on the results of the performance evaluation. (1) Optimize the target data requirement set: Add the data requirement item "interaction frequency of recommender"; (2) Optimize data acquisition and processing strategies: Adjust the intensity of Laplace noise to improve the accuracy of numerical data processing by 5%; (3) Optimize the social network graph: Adjust the weight of the relationship edge strength calculation, increase the weight of "interaction frequency", and improve the accuracy of identifying high-influence customer nodes.

[0083] This specific embodiment resolves the conflict between business data value mining and compliance risk control, and realizes closed-loop data processing that starts from business objectives, incorporates compliance, and protects privacy, directly empowering core businesses such as precision marketing and risk management.

[0084] It should be noted that the sequence numbers of the steps in the above embodiments do not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. In practical applications, all the above possible implementation methods can be arbitrarily combined in a combined manner to form possible embodiments of this application, which will not be described in detail here.

[0085] Based on the differential privacy-based social graph construction method provided in the above embodiments, and based on the same inventive concept, this application also provides a differential privacy-based social graph construction apparatus.

[0086] Figure 6 This is a structural diagram of the social graph construction apparatus based on differential privacy provided in an embodiment of this application. Figure 6 As shown, the differential privacy-based social graph construction device may specifically include a demand decomposition unit 210, a risk assessment unit 220, a privacy processing unit 230, a graph construction unit 240, and an application execution unit 250.

[0087] The requirement decomposition unit 210 is used to obtain the target business requirements and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement. Risk assessment unit 220 is used to conduct risk assessment on the data requirements of the target data requirement set based on preset indicators and generate data collection and processing strategies. The privacy processing unit 230 is used to acquire raw data based on the data acquisition and processing strategy, and to perform privacy processing on the raw data using a differential privacy strategy to obtain the target dataset. Graph construction unit 240 is used to construct a social network graph based on the target dataset; Application execution unit 250 is used to construct business decision rules based on social network graphs and execute target business operations.

[0088] This application embodiment provides a possible implementation, wherein the requirement decomposition unit 210 is further configured to: Based on the mind chain guidance model, the target business requirement is broken down into at least one sub-objective; For each sub-objective, identify at least one associated business metric; Based on each business metric, data requirements for building a social network graph are mapped out, forming a target data requirement set.

[0089] This application embodiment provides a possible implementation, wherein the risk assessment unit 220 is further configured to: Obtain the authorization terms information corresponding to the data requirements of the target data requirement set; Based on the authorization terms information, a mapping matrix is ​​established between the authorization terms information and the data usage. The mapping matrix is ​​used to represent the data usage allowed by each authorization term. Based on the mapping relationship matrix and in accordance with the principle of minimum necessity, the purpose relevance assessment, means minimum assessment and scope applicability assessment are carried out for each data requirement in the target data requirement set to obtain the risk assessment result corresponding to each data requirement. Based on the risk assessment results of all data requirements, a data collection and processing strategy is generated.

[0090] This application embodiment provides a possible implementation, wherein the privacy processing unit 230 is further configured to: Identify the data type of the raw data, which includes numerical data, relational data, and categorical and location data; Add Laplace noise to numerical data; Add noise to the edge weights in relational data, or perform random reconnection of the graph structure; An exponential mechanism is used to fuzzify categorical and location-based data.

[0091] This application embodiment provides a possible implementation, wherein the atlas construction unit 240 is further configured to: Based on the target dataset, the constituent elements required to construct a social network graph are identified. The constituent elements include a set of nodes and a set of relation edges. Each node in the set of nodes corresponds to an entity object, and the node contains the unique identification information and attribute information of the entity object. Based on the data relationships contained in the target dataset, define the relationship edges between nodes in the node set, and determine the type and strength of each relationship edge; wherein, the type of relationship edge includes at least one of business association relationship, behavior association relationship, and attribute similarity relationship; A social network graph is constructed based on the set of nodes, the set of relation edges, and the type and strength of each relation edge.

[0092] This application embodiment provides a possible implementation, wherein the application execution unit 250 is further configured to: Identify target nodes with specific business attributes based on social network graphs; Generate suitable business decision rules based on the business attributes of the target node; Based on business decision rules, trigger and execute the corresponding target business operations.

[0093] This application provides a possible implementation method, wherein the differential privacy-based social graph construction device is further used for: Attribution analysis or A / B testing is performed on the performance of the target business operations to obtain performance evaluation results; Based on the results of the effectiveness evaluation, the target data requirement set, data collection and processing strategies, or social network graphs are iteratively optimized.

[0094] Based on the same inventive concept, embodiments of this application also provide an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the differential privacy-based social graph construction method of any of the above embodiments.

[0095] In an exemplary embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7 The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of this electronic device 300 does not constitute a limitation on the embodiments of this application.

[0096] Processor 301 may be a CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0097] Bus 302 may include a pathway for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0098] The memory 303 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0099] The memory 303 stores computer program code that executes the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the computer program code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0100] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0101] Based on the same inventive concept, embodiments of this application also provide a storage medium storing a computer program, wherein the computer program is configured to execute the differential privacy-based social graph construction method of any of the above embodiments at runtime.

[0102] Those skilled in the art will clearly understand that the specific working process of the systems, devices, and modules described above can be referred to the corresponding process in the foregoing method embodiments. For the sake of brevity, it will not be repeated here.

[0103] Those skilled in the art will understand that the technical solution of this application, or all or part of it, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several program instructions to cause an electronic device (e.g., a personal computer, server, or network device) to execute all or part of the steps of the methods described in the embodiments of this application when running the program instructions. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0104] Alternatively, all or part of the steps of the foregoing method embodiments can be implemented by hardware (such as electronic devices like personal computers, servers, or network devices) associated with program instructions. The program instructions can be stored in a computer-readable storage medium. When the program instructions are executed by the processor of the electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of this application.

[0105] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that within the spirit and principles of this application, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the corresponding technical solutions to leave the protection scope of this application.

Claims

1. A method for constructing a social graph based on differential privacy, characterized in that, The method includes: Obtain the target business requirements, and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement; Based on preset indicators, a risk assessment is performed on the data requirements of the target data requirement set to generate a data collection and processing strategy. Based on the data acquisition and processing strategy, raw data is acquired, and the raw data is then processed for privacy using a differential privacy strategy to obtain the target dataset. Construct a social network graph based on the target dataset; Business decision rules are constructed based on the social network graph, and target business operations are executed.

2. The method according to claim 1, characterized in that, The process of acquiring target business requirements and generating a target data requirement set based on those requirements includes: Based on the mind chain guidance model, the target business requirement is broken down into at least one sub-target; For each of the sub-objectives, at least one associated business metric shall be identified; Based on each of the aforementioned business metrics, data requirement items for constructing the social network graph are mapped to form the target data requirement set.

3. The method according to claim 1, characterized in that, The step of conducting a risk assessment of the data requirements of the target data requirement set based on preset indicators and generating a data collection and processing strategy includes: Obtain the authorization terms information corresponding to the data requirements of the target data requirement set; Based on the authorization terms information, a mapping matrix is ​​established between the authorization terms information and data usage, and the mapping matrix is ​​used to represent the data usage permitted by each authorization term; Based on the mapping matrix and in accordance with the principle of minimum necessity, each data requirement in the target data requirement set is evaluated for purpose relevance, means minimumity, and scope applicability to obtain the risk assessment result corresponding to each data requirement. Based on the risk assessment results of all data requirements, the data collection and processing strategy is generated.

4. The method according to claim 1, characterized in that, The privacy processing of the original data using a differential privacy strategy includes: Identify the data type of the raw data, which includes numerical data, relational data, and categorical and location data; Add Laplace noise to numerical data; Add noise to the edge weights in relational data, or perform random reconnection of the graph structure; An exponential mechanism is used to fuzzify categorical and location-based data.

5. The method according to any one of claims 1 to 4, characterized in that, The construction of the social network graph based on the target dataset includes: Based on the target dataset, the constituent elements required to construct the social network graph are identified. The constituent elements include a set of nodes and a set of relation edges. Each node in the set of nodes corresponds to an entity object, and the node contains the unique identification information and attribute information of the entity object. Based on the data relationships contained in the target dataset, define the relationship edges between the nodes in the node set, and determine the type and strength of each relationship edge; wherein, the type of relationship edge includes at least one of business association relationship, behavior association relationship, and attribute similarity relationship; The social network graph is constructed based on the set of nodes, the set of relation edges, and the type and strength of each relation edge.

6. The method according to claim 5, characterized in that, The step of constructing business decision rules based on the social network graph and executing target business operations includes: Based on the social network graph, target nodes with specific business attributes are identified; Based on the business attributes of the target node, generate the appropriate business decision rules; Based on the aforementioned business decision rules, the corresponding target business operation is triggered and executed.

7. The method according to claim 1, characterized in that, The method further includes: Attribution analysis or A / B testing is performed on the execution effect of the target business operation to obtain the effect evaluation results; Based on the effect evaluation results, the target data requirement set, the data collection and processing strategy, or the social network graph are iteratively optimized.

8. A social graph construction device based on differential privacy, characterized in that, The device includes: A requirement decomposition unit is used to obtain target business requirements and generate a target data requirement set based on the target business requirements; the target data requirement set contains at least one data requirement. The risk assessment unit is used to assess the risk of data requirements for the target data requirement set based on preset indicators and generate data collection and processing strategies. The privacy processing unit is used to acquire raw data based on the data acquisition and processing strategy, and to perform privacy processing on the raw data using a differential privacy strategy to obtain the target dataset. The graph construction unit is used to construct a social network graph based on the target dataset; The application execution unit is used to construct business decision rules based on the social network graph and execute target business operations.

9. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the differential privacy-based social graph construction method according to any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the differential privacy-based social graph construction method according to any one of claims 1 to 7 at runtime.

Images