Business data processing method, apparatus, device, storage medium, and program product

Abstract

The application provides a business data processing method and device, equipment, storage medium and program product, which can be applied to the fields of artificial intelligence technology and financial technology. The method comprises the following steps: in response to newly added business receipt data, classifying and extracting the newly added business receipt data to obtain business element data; performing event analysis on a business event captured in a financial business system to obtain a target customer and a business scenario, and obtaining full matching features of the target customer based on the business element data; in response to the fact that a preset service cluster has an inference model corresponding to the business scenario, taking the inference model corresponding to the business scenario as a target inference model; and implementing a business execution strategy output by the target inference model according to the full matching features of the target customer to obtain business receipt data.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence technology and financial technology technology, specifically to a data processing method, and more specifically to a business data processing method, apparatus, device, storage medium and program product. Background Technology

[0002] In business scenarios such as customer operations, product promotion, and customer relationship maintenance, the financial industry generates a large amount of business receipt data and real-time business event data. These various types of data are scattered and stored in different business subsystems of financial institutions, with messy data formats and a lack of unified organization and structured governance methods.

[0003] Under the current financial business data processing model, various types of business data cannot be formed into a unified and reusable data asset. Customer characteristics are fragmented and incomplete, making it impossible to reverse-engineer and reuse them, resulting in a waste of data and storage resources. Individually trained functional models have low adaptability to business scenarios, and their training results cannot be efficiently and with low loss transformed into practical application capabilities. This leads to the inefficient use of functional models in business scenarios, resulting in poor execution accuracy and low operating efficiency of business operation strategies in financial business systems based on functional models. Summary of the Invention

[0004] In view of the above problems, embodiments of this application provide a business data processing method, apparatus, device, medium, and program product.

[0005] According to a first aspect of this application, a business data processing method is provided, comprising: responding to newly added business receipt data, classifying and extracting the newly added business receipt data to obtain business element data; performing event analysis on business events captured in a financial business system to obtain target customers and business scenarios, and obtaining full-scale matching features of target customers based on the business element data; responding to the existence of a pre-defined service cluster with an inference model corresponding to the business scenario, using the inference model corresponding to the business scenario as a target inference model, and iteratively optimizing the inference model based on the business element data; implementing a business execution strategy that outputs data inference for target customers based on the full-scale matching features through the target inference model, and obtaining business receipt data, wherein the business execution strategy includes at least one of target customer group selection, business product push, and business relationship maintenance.

[0006] According to embodiments of this application, newly added business receipt data is classified and extracted to obtain business element data, including: data cleaning and standardization of newly added business receipt data from at least one of a preset gateway, preset SMS platform, business manager terminal, user terminal, and counter system to obtain a standardized business receipt table; data extraction is performed on the standardized business receipt table to obtain a customer unique identifier indicating a business object, a node unique identifier indicating a business node, a type identifier indicating a business node type, and a numerical character indicating business node evaluation information. The business node type includes at least one of transaction type, reach type, and service type, and the numerical character includes the transaction amount. The data includes at least one of the following: reach time and satisfaction; business lineage data is established based on customer unique identifiers, node unique identifiers, and type identifiers to indicate the mapping relationship between business objects and business nodes, and the sequential relationship between business nodes; the type identifiers and numerical characters are matched to obtain business evaluation data to assess the completion of each business node; the business preference characteristics of business objects are obtained from the business lineage data through a pre-trained label generation model; the preference descriptions corresponding to the business preference characteristics are used as business preference data, and the preference descriptions include at least one of risk preference, channel preference, and transaction preference; the business lineage data, business evaluation data, and business preference data are used as business element data.

[0007] According to embodiments of this application, event analysis is performed on business events captured in a financial business system to obtain target customers and business scenarios. Based on business element data, full-scale matching features of the target customers are obtained, including: subscribing to business data streams from a financial business system including at least one of a business manager terminal, a user terminal, and a counter system; capturing business events in the business data stream using key character recognition technology, where business events include at least one of bill generation, product expiration, product page clicks, and fund transfers; determining the target customer based on the unique customer identifier in the business event message, and determining the business scenario based on the event type identifier in the message; obtaining business lineage data, business evaluation data, and business preference data associated with the unique customer identifier from the business element data; and extracting features from the business lineage data, business evaluation data, business preference data, and business scenarios to obtain full-scale matching features.

[0008] According to an embodiment of this application, in response to the absence of an inference model corresponding to a business scenario in the service cluster, a lightweight standardized model package that matches the business scenario and has been pre-trained is obtained from a preset financial business model asset library. The lightweight standardized model package is pre-trained based on business element data, and the business scenario includes at least one of the following: customer reserve scenario, customer churn handling scenario, product recommendation scenario, and service improvement scenario. The lightweight standardized model package is then subjected to runtime environment verification and containerization to obtain a pre-trained model image. After configuring the pre-trained model image with business services through a preset business service center, it is uploaded to the service cluster as the target inference model. The target inference model is labeled with model codes and corresponding business codes, and the business codes correspond to the business scenarios.

[0009] According to an embodiment of this application, a business execution strategy is output by performing data inference on a target customer based on full-match features using a target inference model. This includes: encapsulating the full-match features in a format adapted to the target inference model to obtain an inference request; inputting the inference request into the target inference model so that the target inference model performs data inference based on the full-match features to obtain a business prediction score corresponding to the business scenario; selecting a business execution template from a preset strategy library based on the business prediction score; filling the business execution template with the target customer's basic information and the strategy object corresponding to the business prediction score; and outputting the business execution strategy.

[0010] According to embodiments of this application, in response to a customer reserve scenario, the business execution strategy is target customer selection, and the business prediction score is used to indicate the similarity between the target customer and the preset customer group label, with the strategy object being the target customer's customer group label; in response to a customer churn handling scenario, the business execution strategy is business relationship maintenance, and the business prediction score is used to indicate the level of maintenance required by the target customer, with the strategy object being the category of discounted products or gifts; in response to a product recommendation scenario, the business execution strategy is business product push, and the business prediction score is used to indicate the target customer's preference for the product to be pushed, with the strategy object being the product to be pushed; in response to a service improvement scenario, the business execution strategy is business relationship maintenance, and the business prediction score is used to indicate the level of service improvement required by the target customer, with the strategy object being a service questionnaire or human communication plan.

[0011] According to an embodiment of this application, the method further includes: acquiring newly added business element data, converting the newly added business element data into a feature wide table, and storing the feature wide table in a sample temporary storage; in response to the total number of feature wide tables in the sample temporary storage exceeding a preset temporary storage threshold or the sample temporary storage reaching a preset running cycle, using the feature wide tables in the sample temporary storage as an incremental sample set to iteratively optimize the inference model, and clearing the sample temporary storage.

[0012] According to an embodiment of this application, determining that a service cluster has an inference model corresponding to a business scenario includes: querying a preset scenario model mapping table based on the business code of the business scenario to obtain the model code that is effective for the business code; sending a service query request for the model code to the business service center to obtain the running status of the model corresponding to the model code; if the running status indicates normal operation, then the service cluster has an inference model corresponding to the business scenario.

[0013] According to a second aspect of this application, a business data processing apparatus is provided, comprising: a data storage module, configured to classify and extract newly added business receipt data in response to new business receipt data to obtain business element data; an event extraction module, configured to perform event analysis on business events captured in a financial business system to obtain target customers and business scenarios, and obtain full-scale matching features of target customers based on the business element data; a service expansion module, configured to use the inference model corresponding to the business scenario as the target inference model in response to the existence of a pre-defined service cluster, and obtain the inference model obtained by iterative optimization based on the business element data; and a strategy implementation module, configured to implement a business execution strategy output by data inference of target customers based on full-scale matching features through the target inference model, and obtain business receipt data, wherein the business execution strategy includes at least one of target customer group selection, business product push, and business relationship maintenance.

[0014] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0015] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0016] According to a fifth aspect of this application, a computer program product is also provided, comprising a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description

[0017] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0018] Figure 1 The illustrations depict application scenarios of business data processing methods, apparatuses, devices, media, and program products according to embodiments of this application.

[0019] Figure 2A flowchart illustrating a business data processing method according to an embodiment of this application is shown schematically.

[0020] Figure 3 This illustration shows a schematic diagram of obtaining full-scale matching features in a business data processing method according to an embodiment of this application;

[0021] Figure 4 This illustration schematically shows the operational relationship of the functional architecture of a financial business system in a business data processing method according to an embodiment of this application;

[0022] Figure 5 This schematic diagram illustrates a structural block diagram of a business data processing apparatus according to an embodiment of the present application;

[0023] Figure 6 A block diagram schematically illustrates an electronic device suitable for implementing a business data processing method according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0027] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0028] In the technical solution of this application, the customer information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0029] In scenarios where personal information is used for automated decision-making, the methods, devices, and systems provided in this application all provide users with corresponding operation entry points for users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process enters the expert decision-making process.

[0030] As used in this paper, the term "model" refers to a model that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs using multiple layers of processing units. A neural network model is an example of a deep learning-based model. In this paper, "model" may also be referred to as a "machine learning model," "learning model," "machine learning network," or "learning network," and these terms are used interchangeably.

[0031] Figure 1 The illustration depicts an application scenario diagram of business data processing according to an embodiment of this application. For example... Figure 1 As shown, application scenario 100 according to an embodiment of this application may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables. For example, a user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send information, etc.

[0032] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be electronic devices such as smartphones, wearable devices, personal computers, intelligent voice interaction devices, smart home appliances, intelligent vehicles, in-vehicle terminals, aircraft, unmanned vending terminals, and extended reality devices. Extended reality devices can include virtual reality devices, augmented reality devices, and mixed reality devices. A client application for the target application can be installed and run on the terminal devices. This target application can include, but is not limited to, financial transaction applications, payment applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, and social media platform software (these are just examples). Furthermore, this application embodiment does not limit the form of the target application, and it can include, but is not limited to, applications, mini-programs, etc., installed on the terminal devices, and can also be in the form of web pages.

[0033] Server 105 can be a server providing various services, such as a backend management server supporting websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services such as cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and basic cloud computing services such as big data. The server can be the backend server of the aforementioned target application, used to provide backend services to the clients of the target application.

[0034] It should be noted that the business data processing method provided in the embodiments of this application can generally be executed by server 105 and / or terminal devices 101-103. Accordingly, the business data processing apparatus provided in the embodiments of this application can generally be disposed in server 105 and / or terminal devices 101-103.

[0035] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0036] Figure 2 A flowchart illustrating a business data processing method according to an embodiment of this application is shown schematically. Figure 2 As shown, the business data processing method 200 according to an embodiment of this application may include steps S210 to S240.

[0037] In step S210, in response to the newly added business receipt data, the newly added business receipt data is classified and extracted to obtain business element data.

[0038] As an example, business receipt data refers to the collective term for various business-related data returned by operational support modules such as financial business systems, operational channels, customer terminals, account manager terminals, and counter systems after completing business actions. These data include business execution result data, outreach result data, transaction processing result data, and customer feedback data. The business actions include, but are not limited to, the preset operational actions of the aforementioned operational support modules and the specific execution actions of the business execution strategies described below. This business receipt data serves as the starting and ending point of the closed-loop operation process in the business data processing method. Based on this, business elements can be generated to drive model iteration, forming a closed loop in business data processing. The classification and extraction process is a standardized process for classifying, cleaning, parsing, and structurally extracting business receipt data from multiple sources, of multiple types, and in multiple formats according to business objects, business nodes, node types, business values, and business preferences. This classification and extraction can transform messy receipts into structured data, providing standard input for subsequent feature generation and model inference. The classification involved can be achieved through labeling and other methods to distinguish categories. The business element data can be a structured core business data set formed after classification and extraction. This data may include information indicating customers, business processes, and node relationships; information indicating amounts, times, and satisfaction levels; and information indicating risks, channels, and preferences. The process of classifying and extracting business element data can be completed by a pre-defined intelligent tagging module, which can assign different tags to different business element data. For example, atomic tags, composite tags, tag features, and tag assets are used to indicate the structured core business data set, thereby constructing a complete business profile of the customer for iterative optimization of the inference model and data reasoning. All the aforementioned business element data, its corresponding tags, features, and other data can be stored in a pre-defined database. This database is a pre-defined, centralized database associated with the financial business system, used to uniformly store various business element data such as business lineage data, business evaluation data, and business preference data, as well as derivative data such as tags and feature wide tables corresponding to each element data. This database can serve as a business feature asset repository for financial institutions. The financial business system of a financial institution can access this database when conducting any related business to quickly extract relevant features of the business, customers, and products.

[0039] In step S220, event analysis is performed on the business events captured in the financial business system to obtain target customers and business scenarios, and full matching features of target customers are obtained based on business element data.

[0040] As an example, a pre-defined real-time inference engine can be used to subscribe to real-time event streams and implementation business messages from a financial business system, thereby capturing various business events occurring in real time within the system. These business events refer to various events occurring in real time within the financial business system, such as customer behavior, transaction actions, status changes, and triggering events, including bill generation, product expiration, page clicks, fund transfers, transaction occurrences, and service evaluations. The target customer refers to the unique customer entity that can be parsed from the business event, which can consist of a customer ID, identification number, and financial business system account. This target customer is not limited to customers of financial institutions but can also include employees of financial institutions. This business scenario belongs to a financial business operation scenario that can be further expanded based on this business event. For example, if the business event is product expiration, the financial business operation scenario could be product recommendation; if the business event is service evaluation, the financial business operation scenario could be service improvement; and if the business event is transaction cancellation, the financial business operation scenario could be customer churn management, service improvement, and product recommendation. The full-scale matching feature refers to the features corresponding to all business element data related to the target customer that can be retrieved from the database, such as feature wide tables and tags corresponding to business element data.

[0041] In step S230, in response to the existence of a pre-defined service cluster with a reasoning model corresponding to the business scenario, the reasoning model corresponding to the business scenario is used as the target reasoning model, and the reasoning model is iteratively optimized based on the business element data.

[0042] As an example, a service cluster refers to a containerized service cluster that deploys and runs various inference models. It provides an inference service environment for various functional inference models, offering model loading, inference computation, and dynamic scaling capabilities. The inference model is a functional model, and there can be multiple functional models. Different functional models can correspond to different business scenarios, such as models for product recommendation, models for maintaining business relationships, and models for selecting target customer groups. Essentially, it is a machine learning prediction model trained on business element data from a database and iteratively optimized using that data. It outputs various inference results such as customer prediction scores, risk levels, preference probabilities, and matching degrees. This inference model is trained by an automated modeling module using business element data from the database and its corresponding labels and features to train a neural network. This neural network can be any functional model architecture with inference capabilities, such as a logistic regression network model architecture or a functional module of a large model.

[0043] In operation S240, a business execution strategy is implemented to perform data inference on target customers based on the full matching features through the target inference model, and business receipt data is obtained. The business execution strategy includes at least one of the following: target customer group selection, business product push, and business relationship maintenance.

[0044] As an example, the target inference model is a service instance of an inference model that matches the current business scenario, is online, and can be called normally. It is a standardized inference unit that can be directly called via API and is used to perform actual inference calculations for the business scenario. This data inference refers to the process by which the target inference model performs real-time calculations, prediction scoring, scenario matching, and strategy matching based on full-scale matching features, thereby deriving a predictive result usable in the business scenario. The business execution strategy is this predictive result directly usable in the business scenario. Its essence is the inference result of the actual inference calculation by the target inference model. It is an automatically generated and directly executable business action that benefits the operation of financial institutions. For example, it includes, but is not limited to, target customer group selection, business product push, and business relationship maintenance. This allows the operation support module of the financial business system to directly implement customer group selection, strategy configuration, channel push, and intelligent operation strategies, thereby improving the efficiency and orderliness of financial institution operations. The implementation of this business execution strategy generates corresponding business feedback data, which includes, but is not limited to, the process data of the specific execution action itself, the result data, feedback data, and evaluation data generated by the specific execution action.

[0045] Based on this, the generation of business events determines the target customer and business scenario. A comprehensive analysis of all existing business element data is performed on the target customer to conduct data inference for that business scenario. Business execution strategies are then implemented for that scenario, resulting in new business receipt data. This leads to the addition of new business element data to the database, enabling a closed-loop processing of business receipt data. Categorizing and extracting the newly added business receipt data transforms multi-source data into business element data with a unified structure, creating a reusable and addable database. This improves the utilization rate of business receipt data and avoids data waste. Binding the inference model to the corresponding business scenario and optimizing and iterating the inference model using real business element data from the database solves the problem of low adaptability between related technical functional models and business scenarios, preventing the efficient and low-loss transformation of training results into practical application capabilities. This improves the fit between the inference model and the business scenario, enhances the performance of the inference model in business scenarios, and consequently improves the execution efficiency and accuracy of various business execution strategies in financial business systems that rely on the inference model.

[0046] According to the embodiments of this application, newly added business receipt data is classified and extracted to obtain business element data, including: data cleaning and standardization of newly added business receipt data from at least one of a preset gateway, preset SMS platform, business manager terminal, user terminal, and counter system to obtain a standardized business receipt table; data extraction is performed on the standardized business receipt table to obtain a customer unique identifier indicating the business object, a node unique identifier indicating the business node, a type identifier indicating the business node type, and a numerical character indicating the business node evaluation information. The business node type includes at least one of transaction type, reach type, and service type, and the numerical character includes transaction amount, etc. The data includes at least one of the following: reach time and satisfaction level; business lineage data is established based on customer unique identifiers, node unique identifiers, and type identifiers to indicate the mapping relationship between business objects and business nodes, and the sequential relationship between business nodes; the type identifiers and numerical characters are matched to obtain business evaluation data to assess the completion level of each business node; the business preference characteristics of business objects are obtained from the business lineage data through a pre-trained label generation model; the preference descriptions corresponding to the business preference characteristics are used as business preference data, and the preference descriptions include at least one of risk preference, channel preference, and transaction preference; the business lineage data, business evaluation data, and business preference data are used as business element data.

[0047] As an example, the financial institution can be any institution with financial attributes, not limited to banks, financial product exchanges, etc. For ease of description, a bank is used as an example. The preset gateway, preset SMS platform, business manager terminal, user terminal, and counter system are actually the various business access and business processing terminals / channel systems of the bank. The gateway can be the access gateway for mobile banking applications. The SMS platform can be an application that depends on the message push system. The business manager terminal is the electronic product used by customer managers, such as a PAD. The user terminal can be the customer's mobile phone, customer's web page, or other application pages. The counter system is the offline branch business processing system, etc.; after executing specific business execution strategies such as marketing push, customer follow-up, and business processing based on the above channels, receipt data can be generated. Data cleaning and standardization processing refers to removing invalid data with missing fields, duplicate reports, and abnormal business logic from the business receipt data, and unifying the field format and data encoding rules of the business receipt data generated by the above channels to form a unified structured business element data. The standardized business receipt table is a structured receipt data table with unified fields and standardized format after cleaning and standardization. It serves as the basic data source for extracting various business elements. The customer unique identifier is a number used to uniquely identify a single customer, such as a customer number or ID card number. As a business object identifier, it can generate corresponding entity tags for the business object. The node unique identifier refers to a single business action in the business execution strategy, such as financial advice, follow-up visits, or transactions. Each single business action is assigned a unique number to distinguish individual business process nodes. The type identifier refers to the code that distinguishes the major categories of business nodes, which can be divided into three categories: transaction, outreach, and service, corresponding to the node attribute classification of the tag. Numerical characters refer to the field content that records the quantitative information of the node, including quantifiable data such as transaction amount, outreach time, and customer satisfaction. This serves as the original basis for various inference models or manual business evaluation. Business lineage data is based on the data association information built on customer and node identifiers. It can include the mapping relationship between customers and business nodes, the sequential relationship between different business nodes, etc., thereby enabling the retrieval of the entire lineage information of a business object, customer, employee, product, etc., based on the database. Business evaluation data consists of data formed by binding node types with corresponding numerical characters. This data quantifies the actual performance of each business node and can be used as evaluation labels. The pre-trained label generation model is an AI model pre-trained based on existing business feedback data. It automatically extracts customer preferences from business element data such as business lineage data and outputs corresponding labels for each business element. Business preference features, preference descriptions, and business preference data are customer preference information inferred from business lineage data by the label generation model, thus obtaining the preference labels for business objects.Using this as a derived tag enables inference models with functions such as product push, target customer selection, and business relationship maintenance to perform corresponding push, selection, and maintenance operations according to the preferences of business objects.

[0048] In a more specific example, new business receipt data is received from application gateways, SMS platforms, customer manager PADs, mobile banking, and counter systems. This data is first cleaned, deduplicated, and formatted to generate a standardized business receipt table. Then, unique customer identifiers, node identifiers, type identifiers, and numerical characters are extracted from the standardized receipts. For transaction-related identifiers, such as wealth management purchases and credit card deductions; for outreach-related identifiers, such as SMS outreach identifiers and application push outreach identifiers; for service-related identifiers, such as customer manager follow-up calls; and for numerical characters, such as matching actual transaction amounts for transaction-related data, recording message delivery times for outreach-related data, and entering customer follow-up satisfaction scores for service-related data. Based on the extracted information, business lineage data regarding customer-node mapping and sequence is established. Simultaneously, type and numerical values ​​are combined to obtain business evaluation data. Then, a pre-trained label generation model is used to output customer risk preferences based on the business lineage data, forming business preference data. Different business preference data can be labeled with different tags, such as conservative or stable; channel preference, such as prioritizing applications or offline counters; and transaction preference, such as high-frequency wealth management or large-amount current accounts. Ultimately, business lineage data, business evaluation data, and business preference data are stored as a whole in the database as business element data.

[0049] Based on this, business evaluation data and business lineage data are obtained using various identifiers, improving the efficiency and accuracy of business receipt data organization. A full-chain business lineage is constructed to achieve full-chain traceability of business element data, improving its traceability and compliance. Multi-dimensional preference features are automatically generated using a pre-trained label generation model, enabling the derivation of preference features and their corresponding labels. Labels serve as the representation of features, and after being stored in the database, they can be directly used for subsequent full-scale feature retrieval and inference model iteration, thereby improving the utilization rate of business receipt data and avoiding resource waste.

[0050] Figure 3 The illustration shows a schematic diagram of obtaining full-scale matching features in a business data processing method according to an embodiment of this application.

[0051] like Figure 3As shown, according to an embodiment of this application, event analysis is performed on business events captured in a financial business system to obtain target customers and business scenarios. Based on business element data, full-scale matching features of the target customer are obtained, including: subscribing to business data streams from a financial business system including at least one of a business manager terminal, a user terminal, and a counter system; capturing business events in the business data stream using key character recognition technology, where business events include at least one of bill generation, product expiration, product page clicks, and fund transfers; determining the target customer based on the unique customer identifier in the business event message, and determining the business scenario based on the event type identifier in the message; obtaining business lineage data, business evaluation data, and business preference data associated with the unique customer identifier from the business element data; and extracting features from the business lineage data, business evaluation data, business preference data, and business scenarios to obtain full-scale matching features.

[0052] As an example, a financial business system refers to the sum of banking business hardware and software systems, including customer manager terminals, user terminals, and counter systems. It serves as the business carrier for mobile banking, counter services, customer manager PADs, etc., and is the source of real-time business events. Business data flow refers to the streaming message data continuously reported by various financial systems in real time, i.e., the original business event flow. Keyword recognition technology is a processing method that uses techniques such as field keyword matching and message rule parsing to filter valid business events from massive data streams, filtering invalid system logs, redundant heartbeat data, and other invalid data. Business events can be real-time customer business behaviors such as bill generation, product expiration, product page clicks, and fund transfers, serving as the trigger source for real-time inference in the model. The standardized data packets transmitting business data from various business systems are messages, which encapsulate fields such as customer number, event type, and business parameters. Customer number, ID number, and account number can serve as unique customer identifiers. The event type identifier is the code in the aforementioned message used to mark the type of business, which is used to automatically match the corresponding operational business scenario. A target customer is a single business customer identified by a unique customer identifier. Essentially, it is a business object and the target for subsequent reasoning and strategy implementation. Full-scale matching features can be a structured feature set formed by retrieving three types of business element data from the database: business lineage data, business evaluation data, and business preference data of the target customer, combined with the current business scenario, and then filtering and extracting them. This set is represented as a wide feature table, which summarizes all types of features in a unified table.

[0053] In a more specific example, a comprehensive financial business system is pre-integrated, comprising customer manager PAD terminals, customer mobile user terminals, and offline counter systems. Each business terminal continuously reports real-time business data streams, which contain a mix of system heartbeat messages, invalid logs, and valid business messages. Using key character recognition technology, filtering by keywords such as "bill," "maturity," "click," and "transfer," the system accurately captures four types of valid business events: bill generation, wealth management product maturity, wealth management page clicks, and large-amount fund transfers. For example, after capturing a wealth management product maturity event message, the system extracts a unique customer identifier from the message fields, such as customer number 622208XXXX, and uses this number to locate the corresponding target customer. Based on the event type identifier embedded in the message, the system automatically matches the customer churn handling business scenario. Subsequently, the system initiates a data retrieval request, pulling the customer's corresponding business lineage data, business evaluation data, and business preference data from the database. Combined with the current churn handling business scenario, irrelevant and redundant fields are removed. Targeted extraction of the customer's asset changes over the past three months, historical wealth management purchases, risk and channel preference information is used to generate the customer's full-scale matching features, which are then provided to the subsequent target inference model.

[0054] Based on this, business events are filtered using key characters to automate data stream cleaning and filtering, improving the effectiveness of mass matching features. Relevant business element data across all dimensions of the target customer is acquired, increasing the conversion rate of business feedback data into usable features, thereby improving business data utilization and avoiding data waste. Full-scale matching features originate from business element data, with each business element corresponding one-to-one with the target customer. This allows the full-scale matching features input into the inference model, combined with historical business execution strategies, to accurately reflect target customer preferences. This not only improves the fit between features and the subsequent inference model but also enhances the accuracy of the inference model's reasoning.

[0055] According to the embodiments of this application, in response to the absence of an inference model corresponding to the business scenario in the service cluster, a lightweight standardized model package that matches the business scenario and has been pre-trained is obtained from a preset financial business model asset library. The lightweight standardized model package is pre-trained based on business element data, and the business scenario includes at least one of the following: customer reserve scenario, customer churn handling scenario, product recommendation scenario, and service improvement scenario. The lightweight standardized model package is subjected to runtime environment verification and containerization encapsulation to obtain a pre-trained model image. After configuring the pre-trained model image with business services through a preset business service center, it is launched to the service cluster as the target inference model, and the target inference model is labeled with model code and business code corresponding to the model code. The business code corresponds to the business scenario.

[0056] Figure 4 This illustration schematically shows the operational relationship of the functional architecture of a financial business system in a business data processing method according to an embodiment of this application.

[0057] exist Figure 4 In the example shown, the financial business model asset repository can be a unified model storage warehouse within a financial institution, used to centrally store various lightweight model packages that have undergone compliance review and pre-training, serving as a unified model asset management carrier. This lightweight standardized model package can be a model file package that has been pre-trained based on business element data within the database, with redundant parameters removed and format standardized. For example, it may contain commonly used algorithm models such as graph neural networks, binary tree architecture, and logistic regression. The automated modeling module can then generate finished inference models based on this. Subsequently, a pre-defined model deployment module performs runtime environment verification. This verification checks whether the model's dependent components and system environment versions match the bank's unified baseline, and identifies issues such as version conflicts and missing files. Containerization encapsulation involves packaging the model and its dependent environment to generate an independent image, resulting in a pre-trained model image. This pre-trained model image refers to a standardized image file that has been packaged and can be directly deployed on the cluster. The business service center can be a service registration center pre-defined by the bank for a unified service registration and scheduling platform, responsible for service registration, address management, and status control. Business service configuration involves configuring online parameters such as service resources, network permissions, health check rules, and interface information. The target inference model is a running instance of the inference model that provides API (Application Programming Interface) inference calls after its image deployment and online deployment. It can be integrated into the real-time inference engine or used independently as a functional model. After the target inference model outputs a business execution strategy, the pre-defined operations support module implements this strategy. During implementation, the operations support module can also use the feature wide table output by the smart tag module to obtain specific implementation information such as the contact information of target customers. The feature wide table of the smart tag module is obtained based on business events from the real-time inference engine. The aforementioned smart tag module, automated modeling module, model deployment module, real-time inference engine, and operations support module can all be part of the financial business system, thus forming a functional closed loop within the financial business system. The model code is a unique identifier for each model image, used to distinguish models of different versions and algorithms. The business code is a code that corresponds one-to-one with a business scenario, which can be used to achieve automatic binding and mapping between scenarios and models.

[0058] In a more specific example, a pre-defined financial business model asset library is accessed to filter lightweight, standardized model packages that match product recommendation scenarios. These lightweight, standardized model packages rely on massive amounts of business element data stored in a database, are pre-trained using a logistic regression algorithm, and undergo parameter simplification and lightweighting to meet the bank's online access standards. Next, runtime environment verification is performed to check if the software programming language version and algorithm dependencies required by the model are consistent with the bank's unified operating baseline, and to identify missing packages and version conflicts. After verification, the model is containerized according to the bank's container specifications, generating an independent pre-trained model image and caching it in an internal image repository. The pre-defined business service center completes the online configuration, allocating CPU and memory resources as needed, configuring network access policies and scheduled health check rules, and deploying the image to the service cluster to generate a callable target inference model. Simultaneously, a unique model code is assigned to this inference service, bound to a dedicated business code corresponding to the product recommendation scenario, and entered into the scenario model mapping table. When subsequent business triggers inference, the system can quickly retrieve the service using the business code. This eliminates the need for developers to perform secondary interface development, enabling rapid online deployment and elastic deployment of the inference model.

[0059] Based on this, reusable business element data is used to pre-train models, saving the time of building from scratch. This allows new models adapted to application scenarios to be deployed quickly, improving the deployment efficiency of inference models. Furthermore, the business element data is based on real business feedback data, which enhances the adaptability of the inference model to business scenarios during pre-training. Unified environment verification and container encapsulation enable environment isolation for inference models targeting different business scenarios, improving the compatibility and stability of various inference models, thereby increasing the accuracy of business execution strategy generation and business adaptability. By binding model codes and business codes one-to-one and storing them in a mapping table, automated matching and invocation of scenarios and models are achieved, reducing the adaptation cost between business scenarios and inference models, and consequently reducing the operational costs of the entire business data processing workflow.

[0060] According to an embodiment of this application, a business execution strategy is output by performing data inference on a target customer based on full-match features using a target inference model. This includes: encapsulating the full-match features in a format adapted to the target inference model to obtain an inference request; inputting the inference request into the target inference model so that the target inference model performs data inference based on the full-match features to obtain a business prediction score corresponding to the business scenario; selecting a business execution template from a preset strategy library based on the business prediction score; filling the business execution template with the target customer's basic information and the strategy object corresponding to the business prediction score; and outputting the business execution strategy.

[0061] As an example, full-match features can be multi-dimensional structured feature data of target customers extracted from the database based on business scenarios. This data can cover business lineage, evaluation, and preference data, serving as the input data source for model inference. The format adapted to the target inference model is the feature encapsulation specification pre-agreed upon by various inference services in the service cluster, including field types, data structures, and message formats. This is uniformly determined when the model goes live. The inference service refers to the complete service system for data inference, including the inference model. The inference request is the call message generated after encapsulating features according to the specification format. It is used to trigger the inference function of the target inference model but does not provide the basis for inference. The business prediction score is the quantitative score output by the model based on the input features, including churn probability, product purchase intention, customer group similarity, etc., corresponding to the model's inference output result. The strategy library is a database that pre-stores execution templates corresponding to various business scenarios, which can be pre-configured and maintained by operations personnel. The business execution template can be a standardized blank strategy template preset according to different business scenarios, with a fixed strategy framework; only the customer information and the strategy content corresponding to the score need to be filled in. The strategy target can be an operational entity determined based on predictive scoring, such as target customer group tags, preferential products, push financial products, and follow-up plans. The business execution strategy is a complete operational plan that can be directly deployed to various channels after the business execution template is filled with information. For example, it may include customer group selection, product push, and customer maintenance content.

[0062] In a more specific example, within a bank's wealth management recommendation service, the acquired full-scale customer matching features are encapsulated into a pre-defined file format according to the wealth management inference service agreement to generate an inference request, which is then sent to the target inference model via API. After inference service computation, a wealth management intention prediction score of 0.86 is output. The system retrieves a dedicated product recommendation execution template from the strategy library, inputs basic information such as customer ID and age, and combines this with the strategy object of the high-scoring corresponding wealth management fund product into the template. This automatically generates a "Push stable fund products to the APP" business execution strategy, which is then directly deployed to the mobile banking channel for real-time inference and automatic strategy generation. The basic information of the target customer may include basic attribute data such as customer ID, name, and card opening branch, taken from the database.

[0063] Based on this, all matching features are uniformly encapsulated into inference requests according to service standards, avoiding call failures caused by format anomalies and effectively improving the stability and derivation effectiveness of the inference model. A pre-defined strategy library and business execution templates are automatically matched and populated by the prediction score to generate standardized business execution strategies, eliminating the need for manual strategy writing, significantly reducing operational preparation time, and improving the targeting and scientific nature of business execution strategies. The quantitative results of the inference model are transformed into implementation strategies that can be distributed across all channels, connecting subsequent channel push and receipt processes, bridging the gap between model inference and business implementation, and realizing the transformation of inference model capabilities into actual business value. This not only improves the utilization rate of business receipt data but also enhances the strategy implementation of financial business systems.

[0064] According to the embodiments of this application, in response to a customer reserve scenario, the business execution strategy is target customer selection, and the business prediction score is used to indicate the similarity between the target customer and the preset customer group label, with the strategy object being the target customer's customer group label; in response to a customer churn handling scenario, the business execution strategy is business relationship maintenance, and the business prediction score is used to indicate the level of maintenance required by the target customer, with the strategy object being the category of discounted products or gifts; in response to a product recommendation scenario, the business execution strategy is business product push, and the business prediction score is used to indicate the target customer's preference for the product to be pushed, with the strategy object being the product to be pushed; in response to a service improvement scenario, the business execution strategy is business relationship maintenance, and the business prediction score is used to indicate the level of service improvement required by the target customer, with the strategy object being a service questionnaire or manual communication plan.

[0065] As an example, in the customer acquisition scenario, the business prediction score output by the inference model represents the similarity between the customers to be screened and the target customer group tags. When the similarity score reaches the standard, the corresponding customer group tags are automatically used as strategy objects to complete the batch selection of target customer groups. The selected potential customers are then uniformly stored in the customer acquisition list. In the customer churn management scenario, the prediction score represents the probability of customer churn and the required maintenance level. The higher the score, the higher the churn risk. The system matches corresponding value financial coupons and exclusive gifts as strategy objects to form customer retention business execution strategies, which are then handed over to account managers via PAD for targeted follow-up and retention. In the product recommendation scenario, the prediction score represents the customer's preference for financial products. Based on the score, suitable funds and fixed-term products are selected as strategy objects, generating a dual-channel product push strategy for customer applications and SMS, and automatically sending the corresponding products. In the service improvement scenario, the prediction score reflects customer service shortcomings and improvement needs. The financial business system can match survey questionnaires and personalized human communication solutions as strategy objects to generate business relationship maintenance strategies, which are then followed up and optimized by operations personnel. Once all the above strategies are implemented, new business receipt data is generated and fed back to the database, achieving a closed loop of business data across the entire chain.

[0066] Based on this, the project addresses the technical challenges of inconsistent execution strategies, low operational efficiency, and the inability to achieve refined, scenario-based automated operations. It binds scoring meanings, strategy targets, and execution methods according to business scenarios, improving the standardization of business execution strategies. This achieves full coverage of business scenarios, automating the execution of all types of operational actions, including selection, push notifications, and maintenance, thereby improving the operational efficiency and coverage of the financial business system. Refined and differentiated operations enable the financial business system to improve strategy conversion rates and customer retention rates, achieving efficient transformation from inference model results to operational implementation.

[0067] According to the embodiments of this application, the method further includes: acquiring newly added business element data, converting the newly added business element data into a feature wide table, and storing the feature wide table in a sample temporary storage; in response to the total number of feature wide tables in the sample temporary storage exceeding a preset temporary storage threshold or the sample temporary storage reaching a preset running cycle, using the feature wide tables in the sample temporary storage as an incremental sample set to iteratively optimize the inference model, and clearing the sample temporary storage.

[0068] As an example, generating new business receipt data will produce new business element data. The storage of these new business element data in the database triggers a database change. Database changes refer to the addition and updating of various business element data, i.e., the strategy implementation generates new receipts and processes them to form new business elements. Once the database changes, the newly added business element data can be obtained. This new business element data can be structured data such as lineage, evaluation, and preference data added to the database after the new business receipts have been categorized and extracted. The feature wide table is the structured dataset integrated by the intelligent tag module according to modeling specifications, which can be used as input for model training. The sample buffer can be a storage unit for temporarily caching incremental feature wide tables, used to accumulate iterative samples. The preset buffer threshold can be a pre-set upper limit for sample data storage capacity. The preset running cycle can be a pre-configured fixed time period, such as daily or weekly, as a trigger condition for periodic iteration. The incremental sample set refers to the training dataset formed by summarizing all feature wide tables in the buffer after the trigger condition is reached. Iterative optimization involves using incremental samples to perform incremental training and parameter updates on the original inference model.

[0069] In a more specific example, in wealth management recommendation and churn warning businesses, new business receipt data is generated daily after various channels execute operational strategies. This data is then converted into new business elements and stored in the database. The financial business system can automatically retrieve the new business element data from the database and convert it into a standardized feature wide table, storing it in a sample temporary storage. Two trigger rules can be configured, such as a sample data volume threshold of 50,000 records and a fixed operating cycle of natural weeks. In one specific embodiment, if the feature wide table data in the temporary storage first exceeds the 50,000-record threshold on Wednesday, the system immediately integrates all data in the storage into an incremental sample set, automatically retrieves the original inference model for incremental iterative training, and clears the temporary storage after the model parameters are optimized. In another specific embodiment, if the total sample volume for the week does not reach the threshold, and the preset cycle expires on Sunday, the system also uses the entire database as incremental samples to complete model iterative optimization. The iterated new model is automatically updated to the model asset library, and subsequent new business events can call the optimized target inference model, thereby realizing a complete mechanism of feedback loop, automatic modeling, and continuous model iteration.

[0070] Based on this, the data from real business data is continuously incrementally iterated to continuously optimize the inference accuracy of the inference model. Based on business data processing operations, a complete business closed loop is formed, including adding new business receipt data, adding new elements, accumulating samples, iterating the model, accurate inference, and optimizing operations. This solves the technical problem of the inference model being disconnected from business from the root, thereby improving the utilization rate of business data and enhancing the inference accuracy and practicality of the inference model.

[0071] According to the embodiments of this application, determining that a service cluster has an inference model corresponding to a business scenario includes: querying a preset scenario model mapping table based on the business code of the business scenario to obtain the model code that is effective for the business code; sending a service query request for the model code to the business service center to obtain the running status of the model corresponding to the model code; if the running status indicates normal operation, then the service cluster has an inference model corresponding to the business scenario.

[0072] As an example, the scenario model mapping table is a pre-established storage table that records the binding relationship between business codes and corresponding effective model codes, serving as an index ledger for scenario-based model lookup. Service query requests can be messages initiated by the financial business system to the business service center to query the instance runtime information corresponding to a specified model code. The runtime status can be determined by marking the model service with three status identifiers: online and running normally, down, or experiencing an abnormal failure.

[0073] In one specific embodiment, a scenario model mapping table is configured, where the product recommendation scenario is bound to business code SC004, and the model code corresponding to the model in use for this scenario is MD202505. When a click event is captured on the product page, business code SC004 is identified, and the scenario model mapping table is automatically retrieved to find the corresponding model code MD202505. A service query request is then initiated to the business service center carrying this model code. If the business service center reports that the processor process, heartbeat, and interfaces of the model instance are all normal and the running status is normal, it is determined that the service cluster has a corresponding inference model, and this model is directly retrieved as the target inference model. If the query result shows that the model is down or in an abnormal state, it is determined that there is no available service, and the process automatically jumps to execute the above-mentioned model encapsulation and image deployment process.

[0074] Based on this, by establishing a fixed binding relationship between business codes, model codes, and scenario model mapping tables, the problem of scenario and model matching errors is avoided from the source, improving the adaptability of the inference model to business scenarios. Before invocation, the running status is queried in advance to identify available services and reduce the probability of inference request errors. If the query is normal, the existing inference model is directly invoked; otherwise, an inference model service is automatically created, achieving automatic splitting of two sets of business logic, realizing a closed loop in business data processing, and improving the automation level and business continuity of the financial business system.

[0075] Based on the above-described business data processing method, embodiments of this application also provide a business data processing apparatus. The following will be combined with... Figure 5 The device is described in detail.

[0076] Figure 5 A schematic block diagram of a business data processing apparatus according to an embodiment of this application is shown.

[0077] like Figure 5 As shown, the business data processing device 500 in this embodiment includes a data storage module 510, an event extraction module 520, a service expansion module 530, and a strategy implementation module 540.

[0078] The data storage module 510 is used to classify and extract newly added business receipt data in response to the data, thereby obtaining business element data. In one embodiment, the data storage module 510 can be used to execute step S210 described above, which will not be repeated here.

[0079] The event extraction module 520 is used to perform event analysis on business events captured in the financial business system to obtain target customers and business scenarios, and to obtain full matching features of target customers based on business element data. In one embodiment, the event extraction module 520 can be used to execute step S220 described above, which will not be repeated here.

[0080] The service extension module 530 is used to respond to the existence of a pre-defined service cluster with a corresponding inference model for the business scenario, and uses the inference model corresponding to the business scenario as the target inference model. The inference model is obtained through iterative optimization based on business element data. In one embodiment, the service extension module 530 can be used to execute step S230 described above, which will not be repeated here.

[0081] The strategy implementation module 540 is used to implement the business execution strategy output by the target inference model based on the full matching features to the target customer, and obtain business receipt data. The business execution strategy includes at least one of target customer group selection, business product push and business relationship maintenance. The strategy implementation module 540 can be used to execute the step S240 described above, which will not be repeated here.

[0082] According to an embodiment of this application, the data storage module 510 performs classification and extraction processing on newly added business receipt data to obtain business element data, including: performing data cleaning and standardization processing on newly added business receipt data from at least one of a preset gateway, a preset SMS platform, a business manager terminal, a user terminal, and a counter system to obtain a standardized business receipt table; extracting data from the standardized business receipt table to obtain a customer unique identifier indicating a business object, a node unique identifier indicating a business node, a type identifier indicating a business node type, and a numeric character indicating business node evaluation information. The business node type includes at least one of transaction type, reach type, and service type, and the numeric character includes... At least one of the following is considered: transaction amount, contact time, and satisfaction level; business lineage data is established based on customer unique identifiers, node unique identifiers, and type identifiers to indicate the mapping relationship between business objects and business nodes, and the sequential relationship between business nodes; the type identifiers and numerical characters are matched to obtain business evaluation data to assess the completion level of each business node; the business preference characteristics of business objects are obtained from the business lineage data through a pre-trained label generation model, and the preference descriptions corresponding to the business preference characteristics are used as business preference data, including at least one of risk preference, channel preference, and transaction preference; the business lineage data, business evaluation data, and business preference data are used as business element data.

[0083] According to an embodiment of this application, the event extraction module 520 performs event analysis on business events captured in the financial business system to obtain target customers and business scenarios. Based on business element data, it obtains full-scale matching features of the target customers, including: subscribing to business data streams of financial business systems including at least one of business manager terminals, user terminals, and counter systems; capturing business events in the business data streams through key character recognition technology, whereby business events include at least one of bill generation, product expiration, product page clicks, and fund transfers; determining target customers based on the unique customer identifier in the message of the business event, and determining business scenarios based on the event type identifier in the message; acquiring business lineage data, business evaluation data, and business preference data associated with the unique customer identifier in the business element data; and extracting features from the business lineage data, business evaluation data, business preference data, and business scenarios to obtain full-scale matching features.

[0084] According to an embodiment of this application, in the service extension module 530, in response to the absence of an inference model corresponding to the business scenario in the service cluster, a lightweight standardized model package that matches the business scenario and has been pre-trained is obtained from a preset financial business model asset library. The lightweight standardized model package is pre-trained based on business element data, and the business scenario includes at least one of the following: customer reserve scenario, customer churn handling scenario, product recommendation scenario, and service improvement scenario. The lightweight standardized model package is then subjected to runtime environment verification and containerized encapsulation to obtain a pre-trained model image. After configuring the pre-trained model image with business services through a preset business service center, it is uploaded to the service cluster as the target inference model, and the target inference model is labeled with model codes and corresponding business codes. The business codes correspond to the business scenarios. Determining that the service cluster has an inference model corresponding to the business scenario includes: querying a preset scenario model mapping table based on the business code of the business scenario to obtain the model code that is effective for the business code; sending a service query request for the model code to the business service center to obtain the running status of the model corresponding to the model code. If the running status indicates normal operation, then the service cluster has an inference model corresponding to the business scenario.

[0085] According to an embodiment of this application, the strategy implementation module 540 performs data reasoning on the target customer based on the full-match features through the target reasoning model, and outputs a business execution strategy, including: encapsulating the full-match features in a format adapted to the target reasoning model to obtain a reasoning request; inputting the reasoning request into the target reasoning model so that the target reasoning model performs data reasoning based on the full-match features to obtain a business prediction score corresponding to the business scenario, and selecting a business execution template from a preset strategy library based on the business prediction score, filling the basic information of the target customer and the strategy object corresponding to the business prediction score into the business execution template, and outputting the business execution strategy; wherein, in response to the business scenario being a customer reserve scenario, the business execution strategy is the target customer... Customer segmentation: Business prediction score indicates the similarity between target customers and preset customer group tags; the strategy object is the target customer's customer group tags. Responding to a customer churn handling scenario, the business execution strategy is business relationship maintenance; the business prediction score indicates the required level of maintenance for target customers; the strategy object is the category of discounted products or gifts. Responding to a product recommendation scenario, the business execution strategy is business product push; the business prediction score indicates the target customer's preference for the product to be pushed; the strategy object is the product to be pushed. Responding to a service improvement scenario, the business execution strategy is business relationship maintenance; the business prediction score indicates the required level of service improvement for target customers; the strategy object is service questionnaires or human communication solutions.

[0086] According to an embodiment of this application, it further includes an iterative optimization unit, which is used to acquire newly added business element data, convert the newly added business element data into a feature wide table, and store the feature wide table in a sample temporary storage; in response to the total number of feature wide tables in the sample temporary storage exceeding a preset temporary storage threshold or the sample temporary storage reaching a preset running cycle, the feature wide tables in the sample temporary storage are used as an incremental sample set to iteratively optimize the inference model, and the sample temporary storage is cleared.

[0087] According to embodiments of this application, any multiple modules among the data storage module 510, event extraction module 520, service expansion module 530, and policy implementation module 540 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the data storage module 510, event extraction module 520, service expansion module 530, and policy implementation module 540 can be at least partially implemented as hardware circuitry, such as field-programmable gate arrays, programmable logic arrays, systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits, or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the data storage module 510, event extraction module 520, service extension module 530, and strategy implementation module 540 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0088] Figure 6 A block diagram schematically illustrates an electronic device suitable for implementing a business data processing method according to an embodiment of this application.

[0089] like Figure 6 As shown, an electronic device 600 according to an embodiment of this application includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory 602 or a program loaded from a storage portion 608 into a random access memory 603. The processor 601 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a dedicated microprocessor. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for executing different steps of the method flow according to an embodiment of this application.

[0090] Random access memory 603 stores various programs and data required for the operation of electronic device 600. Processor 601, read-only memory 602, and random access memory 603 are interconnected via bus 604. Processor 601 executes various steps of the method flow according to embodiments of this application by executing programs in read-only memory 602 and / or random access memory 603. It should be noted that the programs may also be stored in one or more memories other than read-only memory 602 and random access memory 603. Processor 601 may also execute various steps of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0091] According to embodiments of this application, the electronic device 600 may further include an input / output interface 605, which is also connected to a bus 604. The electronic device 600 may also include one or more of the following components connected to the input / output interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube, liquid crystal display, etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card, such as a local area network card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 610 as needed so that computer programs read from it can be installed into the storage section 608 as needed.

[0092] Embodiments of this application also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0093] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include the read-only memory 602 described above, and / or random access memory 603, and / or one or more memories other than read-only memory 602 and random access memory 603.

[0094] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0095] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via the communication section 609, and / or installed from the removable medium 611. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0096] In embodiments of this application, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by processor 601, it performs the functions defined in the system of embodiments of this application. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0097] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0098] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0099] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A business data processing method, characterized in that, The method includes: In response to newly added business receipt data, the newly added business receipt data is classified and extracted to obtain business element data; Event analysis is performed on business events captured in the financial business system to obtain target customers and business scenarios, and full matching features of the target customers are obtained based on the business element data; In response to the existence of a pre-defined service cluster with an inference model corresponding to the business scenario, the inference model corresponding to the business scenario is used as the target inference model, and the inference model is obtained by iterative optimization based on the business element data. Implement a business execution strategy that uses the target inference model to perform data inference on the target customer based on the full matching features, and obtain business receipt data. The business execution strategy includes at least one of the following: target customer group selection, business product push, and business relationship maintenance.

2. The method according to claim 1, characterized in that, The newly added business receipt data is categorized and extracted to obtain business element data, including: New business receipt data from at least one of the following sources—preset gateway, preset SMS platform, business manager terminal, user terminal, and counter system—are cleaned and standardized to obtain a standardized business receipt table. Data is extracted from the standardized business receipt form to obtain a unique customer identifier to indicate the business object, a unique node identifier to indicate the business node, a type identifier to indicate the business node type, and a numerical character to indicate the business node evaluation information. The business node type includes at least one of transaction type, reach type, and service type, and the numerical character includes at least one of transaction amount, reach time, and satisfaction level. Based on the customer unique identifier, the node unique identifier, and the type identifier, establish business lineage data to indicate the mapping relationship between business objects and business nodes, and the sequential relationship between business nodes. Combine the type identifier and the numerical characters to obtain business evaluation data to assess the completion of each business node. The business preference features of the business object are obtained by a pre-trained label generation model based on the business lineage data. The preference descriptions corresponding to the business preference features are used as business preference data. The preference descriptions include at least one of risk preference, channel preference, and transaction preference. The business lineage data, the business evaluation data, and the business preference data are used as business element data.

3. The method according to claim 1, characterized in that, Event analysis is performed on business events captured in the financial business system to obtain target customers and business scenarios. Based on the business element data, the full-scale matching characteristics of the target customers are obtained, including: Subscription includes business data streams from at least one of the financial business systems: business manager terminals, user terminals, and counter systems. Business events are captured in the business data stream using key character recognition technology. These business events include at least one of the following: bill generation, product expiration, product page click, and fund transfer. The target customer is determined based on the unique customer identifier in the message of the business event, and the business scenario is determined based on the event type identifier in the message. Obtain the business lineage data, business evaluation data, and business preference data associated with the customer's unique identifier from the business element data; Feature extraction is performed on the business lineage data, business evaluation data, business preference data, and business scenarios to obtain full-scale matching features.

4. The method according to claim 1, characterized in that, In response to the fact that the service cluster does not have an inference model corresponding to the business scenario, Obtain a lightweight standardized model package that matches the business scenario and has been pre-trained from the preset financial business model asset library. The lightweight standardized model package is pre-trained based on the business element data. The business scenario includes at least one of the following: customer reserve scenario, customer churn handling scenario, product recommendation scenario, and service improvement scenario. The lightweight standardized model package is validated for runtime environment and containerized to obtain a pre-trained model image. After configuring the pre-trained model image with business services through a preset business service center, it is launched to the service cluster as a target inference model. The target inference model is labeled with model code and business code corresponding to the model code, and the business code corresponds to the business scenario.

5. The method according to claim 4, characterized in that, The target inference model performs data inference on the target customer based on the full-match features, and outputs a business execution strategy, including: The full set of matching features is encapsulated in a format adapted to the target inference model to obtain an inference request; The inference request is input into the target inference model, so that the target inference model performs data inference based on the full matching features to obtain a business prediction score corresponding to the business scenario. Based on the business prediction score, a business execution template is selected from the preset strategy library. The basic information of the target customer and the strategy object corresponding to the business prediction score are filled into the business execution template, and the business execution strategy is output.

6. The method according to claim 5, characterized in that, In response to the business scenario being the customer reserve scenario, the business execution strategy is to select target customers, the business prediction score is used to indicate the similarity between the target customer and the preset customer group label, and the strategy object is the customer group label of the target customer; In response to the business scenario being the customer churn handling scenario, the business execution strategy is business relationship maintenance, the business prediction score is used to indicate the required level of maintenance for the target customer, and the strategy object is the category of discounted products or gifts; In response to the business scenario being the product recommendation scenario, the business execution strategy is business product push, the business prediction score is used to indicate the target customer's preference for the product to be pushed, and the strategy object is the product to be pushed; In response to the business scenario being the service improvement scenario, the business execution strategy is business relationship maintenance, the business prediction score is used to indicate the degree of service improvement required by the target customer, and the strategy object is a service questionnaire or a manual communication solution.

7. The method according to claim 1, characterized in that, Also includes: Acquire newly added business element data, convert the newly added business element data into a feature wide table, and store the feature wide table in the sample temporary storage library; In response to the total number of feature wide tables in the sample temporary storage exceeding a preset temporary storage threshold or the sample temporary storage reaching a preset running cycle, the feature wide tables in the sample temporary storage are used as an incremental sample set to iteratively optimize the inference model, and the sample temporary storage is cleared.

8. The method according to claim 4, characterized in that, Determining that the service cluster possesses an inference model corresponding to the business scenario includes: Based on the business code of the business scenario, query the preset scenario model mapping table to obtain the model code that is effective for the business code; Send a service query request for the model code to the business service center to obtain the running status of the corresponding model of the model code. If the running status indicates that it is running normally, then the service cluster has an inference model corresponding to the business scenario.

9. A business data processing device, characterized in that, The device includes: The data storage module is used to respond to newly added business receipt data, classify and extract the newly added business receipt data to obtain business element data; The event extraction module is used to perform event analysis on business events captured in the financial business system to obtain target customers and business scenarios, and to obtain the full matching features of the target customers based on the business element data. The service extension module is used to respond to the existence of a pre-defined service cluster with an inference model corresponding to the business scenario, and to use the inference model corresponding to the business scenario as the target inference model. The inference model is obtained by iterative optimization based on the business element data. The strategy implementation module is used to implement the business execution strategy output by the target inference model based on the full matching features to perform data inference on the target customer, and to obtain business receipt data. The business execution strategy includes at least one of the following: target customer group selection, business product push, and business relationship maintenance.

10. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 8.

12. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 8.

Images